Wybrane publikacje pracowników Katedry, cz. 2

Application of Algorithms Dealing with Time Domain Uncertainty for the Automatic Recognition of Musical Phrases

Kostek Bożena and Szczerba Marek. 102nd Convention of the AES, Preprint No. 4502 (N7), Munich, Germany, March 22-25, 1997.

Recognition of musical phrases requires a specific approach according to a specific structure of the phenomena. Possibility of application of different modifications causes that patterns still perceived by human as one melody can be different in time and frequency domains. Application of parametrization can reduce influence of such modifications, however some of them can still misguide recognition process. To improve recognition results a system based on dynamic time warping algorithm has been proposed. Results of preliminary tests are presented. General conclusions concerning proposed method of automatic recognition of musical phrases are derived and presented.

Application of Chebychev Polynomials to Calculation of the Nonlinear Characteristics of the Digital Waveguide Model of the Organ Pipe

Zieliński Sławomir and Szwoch Grzegorz. 102nd Audio Engineering Society Convention, Preprint No. 4499 (N4), Munich, Germany, March 22-25, 1997

Digital waveguide models of organ pipes can serve as a basis for real-time sound synthesis algorithms. A new technique of estimation of nonlinear function modeling the interaction between the air jet and the resonator of the pipe is proposed. Problems related to application of this technique to match desired spectra of organ pipe sounds are discussed.

Application of Neural Networks to the Recognition of Musical Sounds

Kostek Bożena and Królikowski Rafał. Archives of Acoustics, vol. 22, No. 1, 1997, pp. 27-50.

The aim of the presented work was to train a neural network in order to recognize a class of a chosen musical instrument. As problems related to analysis of sounds are related to human subjective perception abilities, then it seems that such tools of analyses as neural nets should be used for recognition processes. On the other hand, an artificial neural network cannot be trained directly with subsequent samples of a sound, thus the feature extraction procedure is needed at first. Unfortunately, there is no consensus regarding the selection of methods for feature vector extraction. There are few approaches to this task. Some of them are based on the source-signal relationships, on the other hand, the arbitrary choice of sound signal parameters is also possible. In the latter case, a set of parameters extracted both from time and frequency domains is created. The experiment aimed to check whether calculated parameters are sufficient for creating a set of sound patterns used for neural network training. Some neural nets were investigated in the experiment, they were trained with so-called ELEVEN and FOURTEEN vector types. After the learning procedure was executed, other examples of the previously created database (but not seen by the neural network) were presented to neural nets. Results show that NNs (neural networks) are able to generalize information included in feature vectors. However, there are some advantages and disadvantages using the NN as a decision algorithm. The main disadvantage of recognizing musical patterns by the NN is the time consumption of a learning phase. On the other hand, when presenting data to NN inputs, there is no problem with variation of parameters within data, and therefore with data clustering, because a NN has the ability to generalize information during the learning phase. In the paper, an analysis of experimental results will be carried on, and conclusions derived from the performed tests will be presented.

Artificial Neural Network As a Classifier of Musical Instrument Sounds

Kostek Bożena and Królikowski Rafał. Proceedings of the 5th European Congress EUFIT97, Aachen, Germany, September 8-12, 1997.

The rationale of this work was to develop artificial neural network for the musical instrument sound classification. For this purpose feed-forward networks using error back-propagation (EBP) algorithm and the delta learning rule were applied. Tasks related to the creation of feature vectors consisted of musical sound parameters were shortly reviewed. For the selected phases of the training process, graphic presentations of dynamic changes of the training parameters were made. Additionally, the relation between the number of iterations and the maximum admissible value of cumulative cycle error was shown. The effectiveness of identifying new objects by the network in the testing phase was presented. Conclusions were also included.

Automatic Reasoning about Acoustic Data Problems with Preprocessing, Classification and Decision Uncertainty

Kostek Bożena. Procedural Conference of the Intelligent Data Analysis (IDA-95), Baden-Baden, Germany, August 17-19, 1995. Print: [Proc.] The International Institute for Advanced Studies in Systems Research and Cybernetics, Vol.1, pp. 99-103.

There are at least three domains of applications of artificial intelligence algorithms in the domain of acoustics. First of all, it is to recall problems connected to the analysis of musical sound in order to qualify what group of musical instruments the analyzed signal belongs to. The second kind of problems related to classification of data concerns analyses of sounds in order to qualify which musical phrase, melody or even which particular musical piece the analyzed musical signals represent. That kind of analyses might provide decisions in automatic recognition of musical style. There is another kind of acoustic data resulting from the subjective testing procedures. This kind of methodology is often used in acoustics due to the influence of subjective judgments on the quality of perceived sounds. The main purpose of this paper is to discuss problems connected to preprocessing and classification of acoustic data with regard to subsequent decision making. As it is impossible to illustrate all the above mentioned topics in this paper, thus these problems will be shown on the basis of an exemplary experiment related to musical instrument timbre recognition.

Digital Waveguide Modeling of the Organ Flue Pipe

Zieliński Sławomir. Proceedings of the 19th Tonmeistertagung, Stafhalle Karlsruhe, Germany, November 15-18, 1996

The digital waveguide model of the organ flue pipe was developed. This model allows one to synthesize realistic sounds of a flue pipe, including transient states. Moreover, it is simple enough to be run in real-time on a single digital signal processor (DSP).

Feature Extraction Methods for the Intelligent Processing of Musical Signals

Kostek Bożena. 99th Convention of the AES, Preprint No. 4076 (H4), New York, NY, USA, October 6-9, 1995.

The purpose of the study was to find appropriate sound parameters that are to be used for feeding inputs of decision algorithms, such as neural network or rough set-based ones. The quality of chosen parameters was tested statistically and with the use of a neural network algorithm. Experimental results and conclusions are to be shown in the paper. Conclusions on artificial intelligence approach to the automatic recognition of musical timbre were added.

Intelligent Analysis of Musical Databases

Kostek Bożena. 4th International Workshop on Rough Sets, Fuzzy Sets, and Machine Discovery (RSFD-96), Conference Materials, pp. 300-305, Tokyo, Japan, November 6-8, 1996.

A rough set-based analysis has been applied to the analysis of musical databases. For this purpose two exemplary musical databases were constructed. The first database consisted of MIDI files is based on Bach's fugues. Another database contained information on musical timbres. Problems connected to the construction of databases and preprocessing of parameters were discussed. Relationships between parameters included in the constructed databases were shown. A rough set-based system for the recognition of musical phrases was employed for the task of the automatic classification of musical timbres. Experimental results were discussed and conclusions included.

MIDI Database for the Automatic Recognition of Musical Phrases

Kostek Bożena and Szczerba Marek. 100th Convention of the AES, Preprint No. 4169 (E-2), Copenhagen, Denmark, May 11-14, 1996.

Musical Databases. Construction and Analysis

Kostek Bożena, Szczerba Marek and Wieczorkowska Alicja. 19th Tonmeistertagung, Conference Materials, Stadthalle Karlsruhe, Germany, November 15-18, 1996.

Parametric Representation of Musical Phrases

Kostek Bożena and Szczerba Marek. 101st Convention of the AES, Preprint No. 4337 (D-3), Los Angeles, CA, USA, November 8-11, 1996.

The goal of this paper is to present musical phrase in a simplified form in order to examine the relationships between its components. For that purpose two different approaches to the musical phrase analysis, namely musicological and MIDI code-based were reviewed and applied to this work. Consequently, a database was created containing the calculated parameters. The quality of applied methods was checked using the so-called analysis-by synthesis approach, in that way recreation of the original phrase was possible. Some general conclusions concerning automatic analysis of musical phrases were derived and presented.

Parametric Representation of Musical Sounds

Kostek Bożena and Wieczorkowska Alicja. Archives of Acoustics, No. 22, 1, 1997.

The rationale of this research work was to find appropriate sound parameters on the basis of which it is possible to discern musical instrument sounds. A review of parameters used in musical acoustics was carried out focusing on the frequency-domain. Some of parameters were extracted from sound representations. Then, the quality of calculated parameters was tested statistically. Additionally, some discretization methods were applied in order to create so-called feature vectors that are to be used for feeding inputs of decision algorithms. Experimental results and conclusions are shown in the paper.

Rough Set Based Analysis of Computer Musical Storage

Kostek Bożena, Szczerba Marek and Czyżewski Andrzej. ICCIMA97, Brisbane, Australia, Month, 1997.

Two musical databases were constructed: the first one consisting of MIDI files based on Bach's fugues and the second one containing information on timbres of musical instruments sounds. Some parametrization methods were introduced to represent features of musical phrases and of musical sounds. Relationships between parameters included in the constructed databases were studied. The rough set-based algorithm was employed to the task of automatic classification of musical patterns. Some experimental results were shown and discussed.

Rough Set-Based Analysis of Musical Databases

Kostek Bożena and Szczerba Marek. EUFIT96 Conference, Vol. 1, pp. 144-148, Aachen, Germany, September 2-5, 1996.

A comparison of various approaches to the musical sound and musical phrase parametrization was presented. For that purpose four databases were created. These databases consisted of objects related to musical timbre and musical phrase representations. A short review concerning the creation of databases along with algorithmic considerations were presented. A rough set-based learning algorithm was applied in order to analyze the properties of parametrized representations. Results of such tests were presented and conclusions were derived from the experiments.

Soft Set Approach to the Subjective Assessment of Sound Quality

Kostek Bożena. InterSymp97, Baden-Baden, Germany, Month, 1997.

An attempt to assess sound quality basing on the soft set approach was made. For that purpose techniques derived from rough set theories have been implemented. The most important notions of rough set theory have been reviewed. A short description of standard testing methods in acoustic sound quality evaluation is also included. Some exemplary data derived from subjective testing are presented and then processed using non-statistical methods. Conclusions concerning applied approaches to the processing of subjective testing results are presented.

Sound Quality Assessment Based on the Rough Set Classifier

Kostek Bożena. Proceedings of the 5th European Congress EUFIT97, Aachen, Germany, September 8-12, 1997.

The aim of this paper is to present the rough set-based approach to the processing of subjective test results. In such tests the quality of sound is evaluated by listeners. They give their opinion on the overall quality of the sound or they assign certain values to assessed sound attributes. In many acoustic domains there is a need to compare sound samples and on this basis to assess the quality of audio equipment, electroacoustic devices, room acoustics, and recently introduced low bit-rate compression algorithms. As there does not exist a consensus on tested features, thus the problem to assess significance of individual attributes may be solved through the soft computing approach. The results of subjective testing are usually gathered in the decision tables containing sound attributes and experts decisions. Therefore, the rough set method was found suitable to the processing of above data. Results of experiments allowing to find hidden relations between sound attributes and experts overall decision will be presented in the final version of the paper.

Study of Parameter Relations in Musical Instrument Patterns

Kostek Bożena and Wieczorkowska Alicja. 100th Convention of the AES, Preprint No. 4173 (E-6), Copenhagen, Denmark, May 11-14, 1996.

Synthesis of Organ Pipe Sound Based on Physical Models

Czyżewski Andrzej, Kostek Bożena and Zieliński Sławomir. Archives of Acoustics, Vol. 21, No. 2, pp.131-147, 1996.

Problems related to the implementation of physical models based synthesis of organ pipe sound are discussed. A new approach to the physical modeling of organ pipe sound, namely waveguide synthesis is introduced. Results of some experiments with this kind of synthesis are presented. Specific features of presented methods and corresponding applications are quoted. Examples of a computer analysis of both synthesized and musical sounds were presented and compared.

A System for Musical Sound Parameter Database Creation and Analysis

Kostek Bożena and Wieczorkowska Alicja. 102nd Convention of the AES, Preprint No. 4498 (N3), Munich, Germany, March 22-25, 1997.

A concept of the system for creating databases of musical instrument sound parameters has been described. The discretization of real-value databases of musical sound parameters has been done using a choice of discretization methods. The distribution of parameter values has been investigated and visualized. Rough set systems have been applied to checking the importance of particular parameters. Conclusions concerning discretization of real-value parameters have been derived.

Application of Fuzzy Logic and Rough Sets to Audio Signal Enhancement

Czyżewski Andrzej and Królikowski Rafał. chapt. 18 in "Rough Fuzzy Hybridization. A New Trend in Decision-Making", Springer-Verlag, Singapore, pp. 397-409, 1999.

A method of noise reduction, related to spectral subtraction and controlled by intelligent algorithms, is described in the paper. A decision system based on fuzzy logic and rough sets is presented. The engineered inference algorithm exploiting rough sets is also included.

Rough Set Analysis of Electrostimulation Test Database for the Prediction of Post-Operative Profits in Cochlear Implanted Patients

Czyżewski Andrzej, Skarżyński Henryk, Kostek Bożena and Królikowski Rafał. 7th Int. Workshop on Rough Sets, Fuzzy Sets, Data Mining, and Granular-Soft Computing, Ube, Yamaguchi, Japonia, 1999

A new method of examining the hearing nerve in deaf people is presented. It consists in testing deaf people with a speech signal delivered via a microelectrode connected to a current source and attached to the promontory. The current delivered to the electrode is modulated with the speech signal, transposed downwards the frequency scale. A database of patients? data and electrostimulation test results was created, and analyzed using a rough set method in order to find rules allowing prediction of hearing recovery of cochlear implantation candidates.

Computational Approach to Spatial Filtering

Czyżewski Andrzej, Lasecki Jacek and Kostek Bożena. 7th European Congress on Intelligent Techniques and Soft Computing, Aachen, Germany, 1999.

Hearing impaired persons have difficulty in understanding speech in cocktail-party conditions. Spatial filtering may be very helpful for such people. This feature should be applied to the hearing aid, thus the computational complexity of spatial filtering-based algorithm must allow real time implementation. In order to meet this assumption some investigations were made and neural network-based algorithm was proposed. This algorithm is presented in this paper.

Investigating Polynomial Approximation of Spectra of the Pipe

Kaczmarek Andrzej, Czyżewski Andrzej and Kostek Bożena. Archives of Acoustics, vol. 24, No. 1.

A precise method for the determination of the spectral representation of pipe sounds was introduced. The polynomial approximation of the spectral envelope was found to be an effective tool, allowing the study of differences between sounds produced by organ pipes of various types belonging to some selected instruments. The paired comparison subjective testing procedure was applied in order to assess the similarities between sounds synthesized using polynomial smoothed spectra and the original organ sound patterns. The statistical processing of test results revealed that a direct relation exists between the type of organ pipe and the minimum order of the approximating polynomial that can be used to represent the pipe sound spectrum, as determined by the positive opinions of the experts. The applied pipe organ sound recording and processing methods, subjective testing procedures and experiment results are discussed in the paper.

Spatial Filtration of Sound for Multimedia Systems

Kostek Bożena, Czyżewski Andrzej and Lasecki Jacek. IEEE Signal Processing Society 1999 Workshop on Multimedia Signal Processing, Copenhagen, Denmark, 1999.

This paper deals with the problem of receiving of a desired signal in noisy or ?cocktail-party" conditions. This problem is vital in many domains, such as communications, multimedia (multimodal interaction), speech recognition, and psychoacoustics (hearing prostheses). It can be partially solved by classical filtering techniques, however these techniques often introduce distortions into the filtered signal. On the other hand, as it results form experiments performed by the authors, a spatial filtration can be performed based on the Artificial Neural Network (ANN). Such an algorithm was elaborated, and some details concerning its implementation are described. Moreover, results of experiments are presented. These results demonstrate that ANN-based nonlinear filter increases the signal-to-noise ratio and improves speech intelligibility.

Multimedia Database of Musical Instrument Sounds

Kostek Bożena and Suchomski Piotr. 134 Acoustical Soc. of America Meeting, Berlin, 1999.

The presented paper addresses the broad problem of automatically recognizing musical instrument sounds. Many applications for the algorithms dealing with these tasks may be foreseen. Nowadays, with the rapid growth of electronic libraries and databases such as those found on the Internet, the possible application may be to search a musical database for the sounds of chosen instruments or for musical tunes. Therefore a multimedia database was prepared which serves as a source of data to be processed by some intelligent algorithms.

Multimedia Fitting System for Hearing Impaired People

Kostek Bożena, Czyżewski Andrzej and Suchomski Piotr. 3rd World Multiconference on Systemics, Cybernetics and Informatics (SCI'99) and the 5th International Conference on Information System Analysis and Synthesis (ISAS'99), Orlando, 1999.

One of the most important stages in the recovery of hearing impaired people is the choice of n adequate hearing aid. The elaborated Multimedia Hearing Aid Fitting System (MHAFS) is an experimental software that allows to find the characterictics of a hearing aid matching patient's needs and to choose automatically a suitable hearing device. It is planned that this system will be made available in the Internet, so it can be used by anybody who is willing to experience a remote approximate testing of hearing characteristics and receive sounds processed like in some well fitted hearing aids. The key issues related to the engineered system will be presented in the paper.

Assessment of Concert Hall Acoustics Using Rough Set and Fuzzy Set Approach.

Kostek Bożena. Chapter in Rough-Fuzzy Hybridization: A New Trend in Decision-Making. Pal S.K., Skowron A. (Eds.), Springer-Verlag. Singapore.

Many literature refs already exist as to how to carry out the process of correlating objective measurements to subjective impressions of an interior space, but there is not yet any consensus on this still unresolved acoustical subject. Recently, a novel approach to computer assessment of acoustical quality has been made using the soft computing approach. Rough set and fuzzy set theories were used for the purpose of processing both subjective evaluation and objective measurement results.

Noise Reduction in Telecommunication Channels Using Rough Sets and Neural Networks

Królikowski Rafał and Czyżewski Andrzej. 7th International Workshop on Rough Sets, Fuzzy Sets, Data Mining, and Granular-Soft Computing, Ube, Yamaguchi, Japonia, 1999.

A new concept of reduction of non-stationary noise affecting audio signals transmitted in telecommunication channels is proposed. This concept exploits some features of the human auditory system as well as some methods originated from soft computing domain, i.e. rough set-based reasoning and neural processing. The foundations of the engineered method and a description of applied decision algorithms are presented. A number of experiments have been prepared, and some of them have already been carried out. A brief discussion of these experiments' results and some conclusions are also included.

Noise Reduction in Acoustic Signals Using the Perceptual Coding

Królikowski Rafał and Czyżewski Andrzej. 137th Regular Meeting of the Acoustical Society of America, Berlin, Germany, 1999

A new method of noise reduction exploiting some features of the human auditory system is proposed by the authors. The noise suppression is obtained twofold: by uplifting masking thresholds and by keeping noisy components just beneath these thresholds. The foundations of the engineered method are described, and some results of the carried out experiments are briefly discussed in the paper.

Electrostimulation Tests as a Tool in Cochlear Implant Preoperative Diagnostics

Skarżyński Henryk, Czyżewski Andrzej and Kostek Bożena. 137 Acoust. Soc. of Amer. Meeting, Berlin, 1999

The procedures developed at the Institute of Physiology and Pathology of Hearing in Warsaw allow to determine some vital characteristics of the hearing sense that help to make decisions regarding the cochlear implantation. Apart from standard pre-examination procedures a test based on the electrical stimulation via the external auditory canal filled with saline can be performed. In order to evaluate the test results both the dynamics range defined by auditory threshold and uncomfortable loudness level and the Time Difference Limen Test are considered. Moreover in some deaf patients a speech communication was achieved with the use of the ball shaped electrode and the spectral compression of speech signal. In this way, the interpretation of the electrical stimulation test results for the new diagnosed cases was made more reliable.

Noise Reduction in Audio Employing Auditory Masking Approach

Czyżewski Andrzej and Królikowski Rafał. Proc. of the 106th AES Conv., Preprint 4930, Munich, Germany, 08-11 May, 1999.

A new method of noise reduction which exploits some features of the human auditory system is proposed by the authors. The noise suppression is obtained twofold: by uplifting masking thresholds and by keeping noisy components just beneath these thresholds. The foundations of the engineered method are discussed extensively in the paper, and some engineered perceptual noise reduction algorithms are described. The way of introduction of the noise reduction features into an MPEG encoder is demonstrated.

Recognition and Prediction of Music, a Machine Learning Approach

Szczerba Marek. 106th AES Convention, Munich, Germany, 1999.

This paper contains a description of a machine-learning-based system for a recognition and prediction of music. The presented system uses advanced data-mining algorithms: neural networks and rough-sets. The system was applied for two main purposes: recognition of musical structures (phrase, rhythm and harmony) and for prediction of musical elements (melody, rhythm and harmony). The system was optimized for each of the purposes. The problems related to the optimization process are presented. Conclusions concerning application of the machine learning methods to the music domain are derived and included.

Applications of Rough Sets and Neural Nets to Noisy Audio Enhancement

Królikowski Rafał and Czyżewski Andrzej. CD-ROM Proc. of the 7th European Congress on Intelligent Techniques and Soft Computing, Aachen, Germany, 13-16 September, 1999.

A new concept of reduction of non-stationary noise affecting audio signals transmitted in telecommunication channels is proposed. This concept exploits some features of the human auditory system as well as some methods originated from artificial intelligence domain, i.e. reasoning based on rough sets and neural processing. The foundations of the engineered method together with a description of applied intelligent decision algorithms are presented in the paper. A number of experiments have been prepared, and some of them have already been carried out. Hence, a brief discussion on the results of these experiments and some conclusions are also included in the paper. The main focus is put on a comparison between different intelligent methods used to non-stationary noise reduction.

Intelligent Echo and Noise Reduction

Czyżewski Andrzej, Królikowski Rafał, Zieliński Sławomir and Kostek Bożena. Proc. of the 3rd World Multiconference on Systemics, Cybernetics and Informatics (SCI'99) and the 5th International Conference on Information System Analysis and Synthesis (ISAS'99), Vol. 4, pp.234-238, Orlando, USA, 31 July -04 August, 1999.

New concepts of echo cancellation and reduction of non-stationary noise affecting audio signals transmitted in telecommunication channels are proposed. In the both cases, some methods originated form artificial intelligence domain, i.e.: genetic algorithms, neural networks, rough sets are applied. In turn, in the noise reduction method, some features of the human auditory system are presented in the paper. Furthermore, a number of experiments have been carried out, and a brief discussion on some of them is included in the paper.

Noise Reduction in Audio Signals Based on the Perceptual Coding Approach

Czyżewski Andrzej and Królikowski Rafał. Proc. of the IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, pp. 147-150, New Paltz, NY, USA, 17-20 October, 1999.

A new concept of the reduction of noise affecting audio signals transmitted in telecommunication channels is proposed. This concept is exploiting some features of the human auditory system. A strong subjective effect of noise suppression in noisy audio can be obtained by uplifting masking thresholds above the estimated level of noisy components or by reducing this level in such a way that the components be maintained just below masking thresholds. The foundations of the engineered method together with the appropriate algorithms are described in the paper. A brief discussion on the results of carried out experiments and some conclusions are also included in the paper. The main focus is put on perceptual foundations of the noise reduction method.

Echo and Noise Reduction Methods for Multimedia Communication Systems

Czyżewski Andrzej, Królikowski Rafał, Zieliński Sławomir and Kostek Bożena. Proc. of the IEEE Signal Processing Society 1999 Workshop on Multimedia Signal Processing, pp.239-244, Copenhagen, Denmark, 13-15 September, 1999.

New concepts of echo cancellation and reduction of non-stationary noise affecting audio signals transmitted in telecommunication channels are proposed. In the both cases, some methods originated form artificial intelligence domain, i.e.: genetic algorithms, neural networks, rough sets are applied. Moreover, in the noise reduction method, some features of the human auditory system are exploited. A number of experiments have been carried out, and a brief discussion on some of them is included in the paper.

A Method for Echo Cancellation in Audio Signals Using the Genetic Algorithm

Czyżewski Andrzej and Zieliński Sławomir. CD-ROM Proc. of the Joint Meeting, 137th regular meeting of the Acoustical Society of America and the 2nd convention of the EAA: Forum Acusticum - integrating the 25th German Acoustics DAGA Conference, Berlin, Germany, 14-19 March, 1999.

In this paper, a new method of echo cancellation is proposed. This method is based on the use of models of systems causing the echo. Parameters of such models are optimized using the genetic algorithm. The computational cost of the proposed method can be minimized by the application of the correlation function.

New Method of Echo Cancellation

Zieliński Sławomir. Proc of the 8th International Symposium on Sound Engineering and Mastering, pp. 31-34, Gdansk, Poland, 9-11 September, 1999.

New method of echo canceller based on the genetic algorithm is proposed. This method is based on the use of models of systems causing the echo. Parameters of such models are optimized using the genetic algorithm.

Sound Synthesis Using Digital Waveguide Modeling

Zieliński Sławomir. Proc of the 8th International Symposium on Sound Engineering and Mastering, pp. 213-216, Gdansk, Poland, 9-11 September, 1999.

A method of digital waveguide modeling was elaborated at Stanford University about ten years ago. Since that time one can observe still increasing growth of its popularity among the researches and companies producing electronic instruments. In this paper, fundamentals of the digital waveguide modeling will be reviewed. Moreover, exemplary sound examples obtained using this method will be presented.

Computer Techniques in Electrostimulation Testing of Hearing and Hearing Aid Modeling

Skarżyński Henryk, Czyżewski Andrzej Kostek Bożena Szwoch Grzegorz. 3rd World Multiconference on Systemics, Cybernetics and Informatics (SCI'99) and the 5th International Conference on Information System Analysis and Synthesis (ISAS'99), Orlando 1999.

In this paper two aplications of computer techniques in audiology are presented. In the first part of the paper, a new method to examine electrostimulation of structure of the auditory tract, developed in the Institute of Physiology and Pathology of Hearing, is described. The study is dedicated to the problem of an evaluation of the auditory nerve electrical sensitivity in deaf people, assisted by computer technology. A new method is proposed, which enables an assessment of both hearing loss in a given moment of time and the future benefits of the cochlear implant to the patient. In the second part of the paper, a new method of fitting acoustical elements of a hearing aid is proposed. A digital waveguide model of these elements is designed. Next,on the basis of this model computer simulations are performed. It is possible to obtain the desired shape of transfer function of the model by changing the values of its parameters. Resulting a computer simulation dimensions of the physical system can be calculated. This method can be used to design acoustical elements of a hearing aid, having desired acoustical properties. Both applications, although aimed at different group of patient.

Prediction of Post-Operative Profits in Cochlear Implanted Patients Using the Electricostimulation Procedure

Skarżyński Henryk, Czyżewski Andrzej and Kostek Bożena. IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, New Paltz, NY, USA, 1999

The presented research is devoted to the problem of evaluation of the auditory nerve electrical sensitivity in deaf people. In the case of profound hearing impairments the assessment of the degree of hearing loss by using standard acoustic tests such as tonal or vocal audiometry, ABR testing, impedance audiometry, etc. would often conclude in a complete lack of response to an acoustic stimulus in the patient. That is why other diagnostic methods that would enable evalualion of the auditory nerve electrical sensitivity have been designed and introduced to the clinical practice. The method of speech signal transmission to the auditory nerve before cochlear implantation was conceived and tested. This method uses spectral transposition of signal delivered to the external electrode allowing to stimulate the auditory nerve.

Modeling of Acoustics of Hearing Aid Earmold Systems

Szwoch Grzegorz, Kostek Bożena. 137 Acoust. Soc. of Amer. Meeting, Berlin 1999.

This paper addresses problems related to modeling acoustical parts of hearing prostheses. In the case of the behind the ear (BTE) hearing aid sound from the transducer is transferred to the auditory channel. Serving this purpose is the acoustic waveguide of the hearing aid. In more advanced hearing aids, this part is acoustically fitted, according to the patient's needs. This is done experimentally during the fitting process. On the other hand, modeling the waveguide can be based on the physical modeling of acoustical systems. The proposed approach and some preliminary results will be presented in the paper.

Designing Waveguide Elements of a Hearing Aid Using the Physical Modeling Techniques

Szwoch Grzegorz, Kostek Bożena, Czyżewski Andrzej. 106th AES Convention, Munich 1999.

The aim of this paper is to model a desired transfer function of a hearing aid. For this purpose physical modeling techniques were used in order to change parameters of a model in real time. The main features of a system allow one to design waveguide elements of a hearing aid. Such a system may be helpful in the process of fitting some hearing aid elements.

A Novel Approach to Echo Cancellation

Czyżewski Andrzej and Zieliński Sławomir. 106th Audio Engineering Society Convention, Munich, Germany, 1999

In this paper a new method of echo cancellation is proposed. This method is based on the genetic algorithm.

Simulating Acoustics of Hearing Aid Employing Non Linear Signal Filtering and Waveguide Modeling

Szwoch Grzegorz, Kostek Bożena, Czyżewski Andrzej. 108th AES Convention, Paris 2000.

A model of hearing aid is designed and used to perform computer simulations that include signal processing (amplification, filtering and compression) as well as transmitting the sound to the ear by the acoustical waveguide. The method and some results of simulations are presented applicable to the process of fitting the hearing aid to the individual patient's needs.

Determining Influence of Visual Cues on the Perception of Surround Sound Using Soft Computing

Czyżewski Andrzej, Kostek Bożena, Odya Piotr, Zieliński Sławomir. RSCTC'2000, Banff, Canada, 2000.

Contemporary digital video, film or multimedia presentations are often accompanied by the surround sound. Techniques and standards involved in digital video processing are much more developed than concepts underlying creating recording and mixing of the multichannel sound. The main challenge in the sound processing in the multichannel system is to create an appropriate basis for the relating multimodal context of visual and sound domains. Therefore, one of the purposes of experiments is to study in which way and how the surround sound interferes or is associated with the visual context. This kind of study was hitherto carried out when two-channel sound technique was associated with a stereo TV. However, there is not much study done yet that associates surround sound and digital video presented at the TV screen. The main issue in such experiments is the analysis of the influence of visual cues on perception of the surround sound. This problem will be solved with the application of fuzzy logic to the processing of subjective test results.

Influence of visual cues on the perception of surround sound

Czyżewski Andrzej, Kornacki Artur, Kostek Bożena, Odya Piotr, Zieliński Sławomir. 139th Meeting of the Acoustical Society of America, J. Acoust. Soc. Amer., Atlanta, USA, 2000

Contemporary digital video, film or multimedia presentations are often accompanied by the surround sound. Techniques and standards involved in digital video processing are much more developed than concepts underlying creating recording and mixing of the multichannel sound. The main challenge in the sound processing in the multichannel system is to create an appropriate basis for connecting multimodal context of visual and sound domains. Therefore one of the purposes of experiments is to study in which way and how the surround sound interfere or is associated with the visual context. This kind of study was hitherto carried out when two-channel sound technique was associated with a stereo TV. However, there is not much study done yet that associate surround sound and digital video presented at the TV screen. The main issue in such experiments is the analysis of the influence of visual cues on perception of the surround sound. This problem will be addressed in the paper.

Neuro-Rough Control of Masking Thresholds for Audio Signal Enhancement

Czyżewski Andrzej, Królikowski Rafał. Journal of Neurocomputing

The paper addresses the problem of neuro-rough hybridisation applied to non-stationary noise reduction. The goal of the intelligent controller is to estimate the current statistics of corrupting noise on the basis of the analysis of signals taken from telecommunication channel. Thereafter, the noise estimate enables determining the masking threshold levels which allow making the noise inaudible in the audio. Since the implemented decision algorithm requires quantised data, thus the Kohonen?s self-organising maps extended by various distance metrics were used as data quantisers. Some results of the experiments in the domain of non-stationary noise reduction in speech are discussed in the paper.

Expert System for Hearing Aids Fitting

Czyżewski Andrzej, Kostek Bożena, Suchomski Piotr. 108th AES Convention, Paris, France, 2000

The engineered experimental software allows to find the characteristics of a fearing aid matching patients needs and to choose automatically a suitable hearing device characteristics. The key issues related to the engineered application are based on the expert system implementation. This expert system uses both fuzzy logic and rough set processing of analytical data. The principles of the engineered expert system application and some details of the rough set and fuzzy logic implementation will be presented in the paper.

Multimedia Technology Based Orientation System for Visually Impaired People

Czyżewski Andrzej, Kostek Bożena. 4th World Automation Congress, WAC 2000, Maui, Hawaii, USA, 2000

The research performed by authors consisted in multimedia system that aimed at enabling people orientating in their surrounding and to avoid any kind of obstacles and theats. The latter aim may be achieved by an intelligently controlled synthesis of the acoustic field based on the digital image analysis. One of the features of such a system should be the ability to identify the location and to describe the dimension of obstacles in the environment. The idea of perceiving the "sound picture" instead of the visual one brings with itself many issues that are closely related to the research subject discussed in the paper.

An Approach to the Automatic Classification of Musical Sounds

Czyżewski Andrzej, Kostek Bożena. 108th AES Convention, Paris, France, 2000

A study on the automatic classification of musical instrument sounds is presented. For this purpose a large database of musical instrument sounds was built, which consists of both solo and duet stereo recordings. The classification process of musical instrument sounds is done on the basis of some soft computing techniques, such as neural networks. The results of the classification are given as a percentage of musical instrument sounds properly recognized by the system. A discussion of the system efficiency and of its limitations is presented. Conclusions and remarks concerning further development of this study are included.

Automatic classification of musical instrument sounds

Kostek Bożena. 139th Meeting of the Acoustical Society of America, J. Acoust. Soc. Amer., Atlanta, GA, USA, 2000

The aim of the presented study is to show that the process of automatic classification of musical instrument sounds is possible on the basis of a limited number of parameters. However, due to the complexity as well as to the unrepeatable nature of musical sounds, both steady- and transient-states should be taken into account while creating feature vectors. For this purpose a database of musical instrument sounds was built containing various instrument sounds played with a different articulation. Then, this database was used in further experiments consisted of some stages, i.e. preprocessing, parameterization and pattern recognition. The main subject of this study was the optimization of the set of parameters to be included in the feature vectors.

Multimedia Hearing Aids Fitting System

Kostek Bożena, Czyżewski Andrzej, Skarżyński Henryk, Mazur J. 4th World Automation Congress, WAC 2000, Maui, Hawaii, USA, 2000

The application described in the paper is concerned with automatic finding the dynamic characteristics of the hearing aid matching patients needs. The multimedia computer technology makes it practical to organize hearing aid fitting basing on the computer software. Consequently, the proposed method of testing hearing abilities and finding the adequate hearing aid dynamical processing characteristics can be based entirely on multimedia computer technology.The subject of the application is the method of hearing aids fitting employing compressed speech understanding tests in noise and the way of organizing such procedure of hearing aids fit in.

Shift in Localization of Phantom Sound Sources in Surround Sound versus Video Context

Kostek Bożena, Czyżewski Andrzej, Odya Piotr. 21st Tonmeistertagung, Hannover, Germany, 2000

Contemporary digital video, film or multimedia presentations are often accompanied by the surround sound. The visual objects displayed on the screen can affect perception of the phantom sound sources in surround panorama. Therefore, one of the purposes of experiments is to study in which way and how the surround sound interferes or is associated with the visual context. The main issue in such experiments is the analysis of the influence of visual cues on perception of the surround sound. This problem will be solved with the application of fuzzy logic to the processing of subjective test results.

Exploitation of Self-Organising Maps for the Reduction of Non-Stationary Noise in Speech Signals

Królikowski Rafał. ICSC (International Computer Science Conventions) Neural Computation, Berlin, 2000

The paper addresses the problem of reduction of non-stationary noise on the basis of exemplary system architecture. Some innovations were introduced in the system: vector quantisation (VQ) of the altering in time noise statistics, application of Kohonen self-organising maps (SOM) as an intelligent controller for managing VQ, and exploitation of some masking properties of the human auditory system. In the paper, some results of the experiments in the domain of non-stationary noise reduction in speech are also discussed.

Localization of Sound Sources by Means of Recurrent Neural Networks

Królikowski Rafał, Czyżewski Andrzej, Kostek Bożena. The Second International Conference on Rough Sets and Current Trends in Computing, Banff, Kanada, 2000

The issue of localization of sound sources for videoconferencing is discussed in the paper. A new algorithm for estimating speaker locations, based on recurrent neural networks (RNN), is introduced and described. The scheme of experiments carried out in an acoustically adopted chamber, exploiting the engineered method is detailed.

Simulation of the Reverberant Space in the Multichannel Audio Using the Convolution Method

Czyżewski Andrzej, Kornacki Artur, Szwoch Grzegorz, Kostek Bożena. 17th International Congress on Acoustics, Rome, Italy, 2001

The convolution method is commonly used to simulate the reverberant space by convolving monophonic or stereophonic sounds with the impulse responses of the room.In this paper,application of this method to the multichannel audio is proposed. The impulse responses of the real room were recorded.Each of the audio channels was obtained using the convolution of the adequate room impulse response with monophonic source sound.The results of the convolution were then combined and encoded as the multichannel surround audio in the format 5.1. The time and spectral analyses of the resulting sounds,as well as the listening tests were performed.The results of these experiments are presented and discussed in the paper. The presented method allows one to simulate the acoustical conditions of the room where the monophonic audio was acquired. Possible applications of this method include advanced Internet teleconferencing in which the bandwidth requirements may be decreased by transmitting only monophonic sounds and the impulse responses of the room instead of the whole multichannel audio.

Automatic Identification of Sound Source Direction Based on Neural Networks

Czyżewski Andrzej, Królikowski Rafał. 142nd Meeting of the Acoustical Society of America, J. Acoust. Soc. Amer., Fort Lauderdale, USA, 2001

In this paper a method for automatic detection of sound source was studied. Both standard feed-forward- and recurrent neural networks were employed in that method. Comparison of the results obtained is given. Conclusions are also derived.

Neural Networks Applied to Sound Source Localization

Czyżewski Andrzej, Królikowski Rafał, Kostek Bożena. 110th Audio Engineering Society Convention, Amsterdam, Netherlands, 2001

The primary aim of this paper is to show that it is possible to localise the direction of the incoming acoustical signal based on the neural network trained for that purpose. Consequently, the automatically localised acoustical signal may be attenuated if it obscures the desired target sound. A set of parameters was formulated in order to localise target source and unwanted signals. In order to process acoustical signals incoming from various directions at the same time the neural network-based system was designed and implemented. The feature extraction method is thoroughly discussed, the training process is described and recently obtained results are discussed.

Acquisition of Acoustic Signals Assisted by Recurrent Neural Networks

Czyżewski Andrzej, Królikowski Rafał. 17th International Congress on Acoustics, Rzym, Włochy, 2001

The issue of localisation of sound sources for videoconferencing is addressed in the paper, where a new method for estimating speaker locations is introduced. It is based on exploitation of temporal relationships between signals received by an array of microphones, and thereby recurrent neural networks are employed. Additionally, a parametrisation of the time-domain audio signals prior to the neural processing is performed. Some of the results of the experiments are briefly presented in the paper.

Determining Influence of Visual Cues on the Perception of Surround Sound Using Soft Computing

Czyżewski Andrzej, Kostek Bożena, Odya Piotr, Zieliński Sławomir. Series: Lecture Notes in Computer Science, vol. 2005, Springer-Verlag, 2001

The main challenge in the sound processing in the multichannel system is to create an appropriate basis for the relating multimodal context of visual and sound domains. Therefore, one of the purposes of experiments is to study in which way and how the surround sound interferes or is associated with the visual context. This kind of study was hitherto carried out when two-channel sound technique was associated with a stereo TV

Digital Waveguide Models of the Panpipes

Czyżewski Andrzej, Jaroszuk Jarosław, Kostek Bożena. ISMA'2001, Perugia, Italy, 2001

The principal aim of this paper is to present a digital waveguide model of the Panpipes. For the efficient modeling of the Panpipes instrument its structure and its physics were studied and thoroughly discussed. The acquired knowledge was then used during the construction of the model. In this context principles of the digital waveguide modeling of woodwind instruments are shortly reviewed. Because of the simplicity of designing the digital waveguide as a set of delay lines and scattering junctions the model can be easily implemented to a digital signal processor. In the paper two digital waveguide models of the Panpipes instruments were presented. They differ from each other by their complexity. This was due to examining the influence of decreasing the complexity of the model on the synthetic sound quality. The performed subjective tests resulted in showing that introduced simplifications in digital waveguide models reveal no noticeable influence on the sound quality. A comparison between synthetic and real Panpipes sounds was made. The results of both subjective tests and objective analyses obtained using engineered models of Panpipes are also included in the paper. Conclusions are derived.

Waveguide Modeling of Ancient, Japanese Musical Instruments

Czyżewski Andrzej, Kostek Bożena, Zieliński Sławomir. ISMA'2001, Perugia, Italy, 2001

Problems related to the implementation of physical modeling-based synthesis of two traditional Japanese instruments are discussed. Examples of computer analyses of sounds of shakuhachi and koto are presented. On the basis of these analyses some assumptions concerning waveguide models were made. Physical modeling principles of musical instrument sounds generation were also shortly reviewed. Main differences in modeling wind and string instruments were highlighted. The process of constructing models of these two musical instruments was explained. A short discussion concerning problems occurred while creating such models was given. Some general conclusions concerning real-time implementation of the digital waveguide models were also included.

Neuro-Rough Control of Masking Thresholds for Audio Signal Enhancement

Czyżewski Andrzej, Królikowski Rafał.

The paper addresses the problem of neuro-rough hybridisation applied to non-stationary noise reduction. The goal of the intelligent controller is to estimate the current statistics of corrupting noise on the basis of the analysis of signals taken from telecommunication channel. Thereafter, the noise estimate enables determining the masking threshold levels which allow making the noise inaudible in the audio. Since the implemented decision algorithm requires quantised data, thus the Kohonen's self-organising maps extended by various distance metrics were used as data quantisers. Some results of the experiments in the domain of non-stationary noise reduction in speech are discussed in the paper.

Dereverberation Based on the Genetic Algorithm

Czyżewski Andrzej, Zieliński Sławomir. 17th International Congress on Acoustics, Rome, Italy, 2001

In this paper, a new method of echo cancellation is proposed applicable to some telecommunication systems. This method is based on the application of the reverse model of a system causing echo. Parameters of such a model are optimized using the genetic algorithm. Some exemplary results of echo cancellation obtained with the use of the proposed method are discussed.

Encoding Spatial Information for Advanced Teleconferencing

Czyżewski Andrzej, Królikowski Rafał, Kostek Bożena. 19th International AES Conference, Schloss Elmau, Germany, 2001

The aim of this paper is to show a system that enables automatic identification of a sound source position in noisy acoustical conditions with a considerable accuracy. Automatic detection of sound source in such an acoustical environment is much needed in advanced teleconferencing. The approach shown in the paper is based on Artificial Neural Networks (ANNs) used for automatic sound localisation. Both standard feed-forward ANNs and Recurrent Neural Networks (RNNs) are employed for that purpose. Comparison of the results obtained, based on both types of ANNs, is also given. Conclusions are derived and shown.

The Internet Sound Restoration Service Based on the Perceptual Denoising Method

Czyżewski Andrzej. 20th Audio Eng. Soc. International Conference, Budapest, Hungary, 2001

The Internet service was launched intended to on-demand restoration and publishing of audio content related to world's cultural heritage. A special way of acquiring, processing and publishing archive recordings was conceived in order to ensure a proper dissemination of the proposed service and its long-term maintenance. The sound enhancement method underlying the system operation employs the extended perceptual coding of audio material allowing for simultaneous noise reduction and sound compression. Moreover, the non-linear predictor employing neural networks was applied to the detection and removal of impulse distortions. The system is still in the development phase, thus both: system features implemented already and technical assumptions related to its further development are presented in the paper.

Discovering the Influence of Visual Stimuli on The Perception of Surround Sound Using Genetic Algorithms

Czyżewski Andrzej, Kostek Bożena, Odya Piotr, Smolinski T. 19th International AES Conference, Schloss Elmau, Germany, 2001

The paper contains a description of experiments that aim to determine visual cue influence on the perception of spatial sound. Earlier stage of the carried out experiments showed that there exists a relationship between the perception of video presented in the screen and sound signals reproduced in a surround system. However, this relationship is dependent on the type of audio-visual signals. Thus a series of subjective tests has been performed on dozens of experts in order to discover these dependencies. The main issue in such experiments is the analysis of the influence of visual cues on the perception of the surround sound. This problem is solved with the application of genetic algorithms to the processing of subjective test results. Conclusions concerning the complexity of the investigated problem are included.

Applications of Neural Networks and Perceptual Masking to Audio Restoration

Czyżewski Andrzej. Journ. of New Musical Reseach, vol. 30, No. 4

Applications of learning algorithms to the restoration of recordings are presented. Attention is paid to the usage of artificial neural networks as a decision system determining which components of an input signal are valid and which ones are unwanted. It provides a basis for the parasitic impulse detection and for the interpolation of lost signal intervals. Such an approach enables also an efficient noise reduction employing the extended perceptual coding algorithm. The proposed algorithms are described briefly in the paper, obtained results are discussed and some general conclusions concerning the application of soft computing and perceptual masking to sound restoration are added.

Problems Related to Surround Sound Production

Kornacki Artur, Kostek Bożena, Odya Piotr, Czyżewski Andrzej. 110th AES Convention, Amsterdam, Netherlands, 2001

The problem of production of recordings designated for sound surround systems becomes a vital problem in sound technology. Existing standards of surround systems allow for reproduction of spatial sound. However, there are no consistent recommendations as to which microphone and mixing technique could be used in specific situations. For the purpose of research presented in this paper several microphone techniques were used for recordings of a quartet playing classical music. The mixing results in two-channel excerpts and several multichannel ones designated for 5.1 reproduction system. Then, in order to find the most preferable recording technique these excerpts were used in subjective tests.

Multimedia Techniques Applied to Health Care Procedures- Hearing Aid Fitting expert System

Kostek Bożena, Czyżewski Andrzej. 46 Internationales Wissenschaftliches Kolloquium, Ilmenau, Germany, 2001

In this paper an exemplary implementation of the complex multimedia system in the domain of the health care and its integration to the user environment is shown. The engineered Multimedia Hearing Aid Fitting Expert System is an experimental software program that allows finding automatically characteristics of a hearing aid matching patients needs. The fitting of the hearing aids is based either on classical methods that use audiometric test results or on loudness scaling principles. All these methods are based on artificial test signals. However, the fitting of hearing aids should be performed on the basis of testing speech understanding in noise. A satisfying reliability of these tests may be achieved through the use of modern computer technology, properly calibrated. The principles of the engineered software application, some details of the calibration process, and results of the experiments will be presented in the paper.

Expert system for Musical Style Recognition

Kostek Bożena. International Workshop: Human Supervision and Control in Engineering and Music, Stadthalle Kassel, Germany, 2001

In this overview some concepts concerning sound engineering, computer music and human supervision are presented. Multimodal-computer interactions consist in, among others, collecting and intelligent searching music related-information. Some concepts related to the author's experience will be presented. Key findings in sound engineering allow recording music in a natural way. Computers can be employed as both Internet sites collecting music-related data and as algorithmic tools that enable musicians to find needed information. They allow analyzing a given melody, modify it in musically sensible ways, mimic the human way of composing, etc. Human supervision is needed at both stages. The quality of recording cannot be assigned otherwise than subjectively. Organizing a computer site containing music-related information needs also the supervising of the future user. Developing artificial intelligence algorithms and designing ergonomic user interfaces is also a task for a human supervisor.

Wavelet-based automatic recognition of musical instruments

Kostek Bożena, Żwan Paweł

The objective of the present work is to automatically extract information from monophonic sounds. This process consists of several stages, namely, preprocessing, parameterization, and classification. This paper shows a thorough study on the wavelet-based parameterization of musical instrument sounds and automatic recognition by means of artificial neural networks (ANNs). First, an engineered method of pitch detection is presented and exemplified by several analyses. A short discussion on error associated with automatic pitch tracking is also included. Then, examples of time-frequency analyses of various musical instrument groups are presented. The analyses are performed employing a database containing musical sounds recorded at the Sound and Vision Engineering Department, Technical University of Gdansk. On the basis of such analyses a set of parameters is derived. Feature vector properties are then discussed. For that purpose Fisher statistics is used. It allows checking the separability between musical instrument pairs. In addition, for the purpose of automatic recognition of musical instrument groups artificial neural networks are used. Various structures and training methods of the ANNs are examined. Exemplary results obtained in the carried out investigations are provided and analyzed. Concluding remarks concerning further development of such experiments are also included in the paper.

Management of Musical Data

Kostek Bożena. International Workshop: Human Supervision and Control in Engineering and Music, Stadthalle Kassel, Germany, 2001

In this overview some concepts concerning future perspectives of transdisciplinary research will be presented. There are many problems related to the management of musical data that are not solved up to now. These problems are being extensively developed within the Music Information Retrieval field now. Topics that should be addressed within the scope of this discussion, but not limited to, are as follows: the problem of automatically classifying musical instrument sounds and musical phrases/styles, music representation and indexing, estimating similarity of music using both perceptual and musical criteria, problems of recognizing music using audio or semantic description, building up musical databases, evaluation of MIR systems, intellectual property right issues, user interfaces, issues related to musical styles and genres, language modeling for music, user needs and expectations, auditory scene analysis, gesture control over musical work, etc. Some of these topic are covered by the MPEG 7 standardization process, which describe the multimedia content data that will support some degree of interpretation of the information meaning, which can be passed onto, or accessed by, a device or a computer code (MPEG-7)

Audio Material Extraction from the Internet Databases

Kostek Bożena. 46 Internationales Wissenschaftliches Kolloquium, Tagungsband, 2001

The paper will outline the problems related to automatic search for audio material. The aim of this paper is to show how to automatically recognize individual musical instrument sounds contained in the Internet sites or multimedia databases. This feature is highly needed in today's Internet browsers. In order to recognize musical instruments properly several stages are needed, namely preprocessing, parameterization, and the actual recognition/classification process. The classification process of musical instrument sounds can be done by means of soft computing techniques that use learn-and-test approach. The main principles of methods for the automatic recognition/classification of musical instrument sounds developed and tested at the Sound & Vision Engineering Department, Technical University of Gdansk will be described. Key challenges in the multimedia technology devoted to this problem will be also presented.

In Search for Surround Sound Recording Techniques

Kostek Bożena, Czyżewski Andrzej. ISMA'2001, Perugia, Italy, 2001

The existing and recently introduced standards of surround systems allow for reproduction of spatial sound in almost any room conditions. The vital concern of sound production for surround systems is the number of microphones, their positioning, proportion between direct sound, early reflections and the reverberation, artificially added delays, etc. The proper solution of such problems may result in creating spatial impression that is comparable to the live music perception. However this kind of a study should address some of the questions related to surround sound production. The broader aim is to establish recommendations as how to produce recordings of classical music designated for sound surround systems in specific acoustical conditions and then to reproduce it properly. This paper shows a study in which several microphone techniques were used for recordings of classical music in two auditory halls having different acoustical properties. Based on these recordings and various mixing techniques two channel stereo excerpts and some multichannel ones were produced. The latter were encoded in 5.1 multichannel format. The extensive subjective tests were performed employing a group of sound engineers and students in order to find the most preferable recording techniques. The listening tests were first performed employing excerpts obtained for each room separately, then the best production was compared for two rooms. The subjective tests were carried out in the same listening room equipped with the 5.1 surround reproduction system. In the paper results of such a comparison tests are shown. The methodology of carrying out subjective tests is presented. The discussion of obtained results and some conclusions are also included.

Representing Musical Instrument Sounds for Their Automatic Classification

Kostek Bożena, Czyżewski Andrzej. J. Audio Eng. Soc., vol. 49, No. 9

A study of the automatic classification of musical instrument sounds is presented. For this purpose a database of musical instrument sound parameters was built which consists of musical instrument recordings and their parametric representations. The parameterization process was conceived and performed in order to find significant musical instrument sound features and to remove redundancy from the musical signal. Classification experiments of musical instrument sounds were performed with neural networks allowing a discussion of the efficiency of the feature extraction process and its limitations. Conclusions and remarks concerning further development of this study and its relation to the current MPEG-7 standardi-zation process are included.

Automatic Recognition of Musical Instrument Sounds - Further Developments

Kostek Bożena, Czyżewski Andrzej. 110th Audio Eng. Soc. Convention, 110th Audio Eng. Soc. Convention, 2001

Discussion on the subject of retrieval of musical data from Internet or multimedia databases, which is carried out now for some time does not successfully reach its final stage of application. There are still many problems related to the subject of automatic recognition of music or musical instrument sounds that cannot be easily solved. Especially important is to find adequate parameters of musical signal based on time and frequency and/or wavelet analyses. Proposed feature vectors were derived on the basis of the constructed databases that contain recorded musical sounds. The presented study shows methods of automatic identification of musical instruments based both on classical statistical and soft computing approaches. They were used then to classify musical instruments. A set of results obtained in the carried out investigations is provided and analyzed and concluding remarks are included in the paper.

Internet-Based Automatic Hearing Assessment System

Kostek Bożena, Czyżewski Andrzej, Skarżyński Henryk, Kochanek K. 46 Internationales Wissenschaftliches Kolloquium Ilmenau, Ilmeanu, Germany, 2001

The aim of this paper is to show the new media application to the domain of health care. In the paper the Internet-based system that allows for automatic testing of hearing is described. Hearing impairment is one of the fastest growing diseases of modern society. Therefore it is very important to organize mass screening tests to identify people suffering from this kind of impairment. The described application provides a test that uses automatic questionnaire analysis, audiometric tone test procedures, and assesses speech intelligibility in noise. When all the testing is completed, the system automatically analyzes the results for each person examined. Based on the number of incorrect answers, the decision is made automatically by the expert system: does the person have normal hearing or does he or she have hearing problems and requires to be examined in one of the consulting centers? Those whose hearing impairment is confirmed are referred to treatment in rehabilitation centers. All these centers are connected via the Internet and are provided with special distributed database access allowing them to automatically register and track the patient discovered during the remote screening.

A method for the automatic hearing aid fitting employing speech in noise

Kostek Bożena, Czyżewski Andrzej. 142nd Acoustical Soc. of America Meeting, Fort Lauderdale, USA, 2001

Some limitations of the hearing aid fitting process are discussed. The classical procedures in this process are based on audiometric test results and/or the loudness scaling method employing artificial test signals. However, the fitting of hearing aids should be also performed on the basis of testing speech understanding in noise, because this is much closer to the real life conditions. A satisfying reliability of these tests may be achieved through the use of modern computer technology with an application of a properly calibrated sound system. A new strategy applicable to fitting prostheses was developed. It allows finding automatically characteristics of a hearing aid matching patients needs. The principles of the fitting method employing fuzzy reasoning, and some results of the experiments will be presented in the paper.

Automatic Recognition of Musical Instrument Sounds

Kostek Bożena. ICA'2001, Rome, Italy, 2001

The presented study aims at the possibility of automatic identification of musical instruments based on signal processing and some intelligent decision techniques. This study points out at automatic retrieval of musical sounds from Internet databases. Several stages should be performed before the actual recognition process takes place. Especially important is to find adequate descriptors of musical sounds. Appropriate sound parameters are to be used for feeding inputs of decision algorithms. They should be well related to sound characteristics, both objectively measured and subjectively perceived. Proposed feature vectors are derived on the basis of thorough examination of sound analysis results. Parameters are looked for in the frequency and time-frequency domains. A discussion concerning the choice of parameters that might be contained in the feature vectors is also included. An expert system based on some classification methods, both classical and soft computing ones is used for automatic classification purposes. Exemplary results obtained in experiments and derived conclusions are included in the paper.

Employing Fuzzy Logic and Noisy Speech for Automatic Fitting of Hearing Aids

Kostek Bożena, Czyżewski Andrzej. 142nd Acoustical Soc. of America Meeting, Fort Lauderdale, USA, 2001

In this paper some limitations of the hearing-aid fitting process are discussed. In the fitting process, an audiologist performs tests on the wearer of the hearing aid, which is then adjusted based on the results of the test, with the goal of making the device work as best as it can for that individual. Traditional fitting procedures employ specialized testing devices which use artificial test signals. Ideally, however, the fitting of hearing aids should also simulate real-world conditions, such as listening to speech in the presence of background noise. Therefore, more satisfying and reliable fitting tests may be achieved through the use of multimedia computers equipped with a properly calibrated sound system. We have developed a new automatic system for fitting hearing aids. It employs fuzzy logic. In this process, a computer makes choices for adjusting the hearing aid's settings by analyzing the patient's responses and answering questions with replies that can lie somewhere between a simple "yes" or "no." This paper will describe the method and present some results of the experiments conducted to test the system.

Localization of Sound Sources by Means of Recurrent Neural Networks

Królikowski Rafał, Czyżewski Andrzej, Kostek Bożena. Series: Lecture Notes in Computer Science, vol. 2005, Springer-Verlag, 2001

The issue of localization of sound sources for videoconferencing is discussed in the paper. A new algorithm for estimating speaker locations, based on recurrent neural networks (RNN), is introduced and described. The scheme of experiments carried out in an acoustically adopted chamber, exploiting the engineered method is detailed.

Prediction of the Reverberation Time in Rectangular Rooms with Non-Uniformly Distributed Sound Absorption

Neubauer R., Kostek Bożena. Archives of Acoustics, vol. 26, No. 3

The aim of this paper is first to review the best known reverberation time formulae and then to show that they cannot predict the reverberation time accurately in cases mostly encountered in practice, where the sound field is not diffuse. Introducing a correction to the Fitzroy's formula allows predicting better the reverberation time in the case of non-uniformly distributed sound absorption. Comparison of calculation results obtained on both the basis of classical equations and the new time reverberation formula introduced is shown and conclusions are drawn.

Determining the influence of visual stimuli on the peception of surround sound using data mining algorithms

Odya Piotr, Czyżewski Andrzej, Kostek Bożena, Smolinski T. 142nd Meeting of the Acoustical Society of America, J. Acoust. Soc. Amer., Fort Lauderdale, USA, 2001

A short description of experiments that aim to determine visual cues influence on the perception of spatial sound is provided in the paper. The earlier stage of the carried out experiments showed that there exists a relationship between the perception of video presented in the screen and sound signals reproduced in a surround system. However, this relationship is dependent on the type of audio-visual signals. Thus a series of subjective tests has been performed on dozens of experts in order to discover these dependencies. The main issue in such experiments is the analysis of the influence of visual cues on the perception of the surround sound. This problem is solved with the application of genetic algorithm and rule searching mechanism to the processing of subjective test results. Some results and conclusions concerning the complexity of the investigated problem are included.

Determination of Influence of Visual Cues on Perception of Spatial Sound

Odya Piotr, Czyżewski Andrzej, Kostek Bożena. 110th Audio Eng. Soc. Conv., Amsterdam, Netherlands, 2001

The paper contains a description of experiments that aim to determine visual cue influence on the perception of spatial sound. Earlier stage of the carried out experiments showed that there exists a relationship between the perception of video presented in the screen and sound signals reproduced in a surround system. However, this relationship is dependent on the type of audio-visual signals. Thus a series of subjective test has been performed on dozens of experts in order to discover these dependencies. The main issue in such experiments is the analysis of the influence of visual cues on the perception of the surround sound. Conclusions concerning the complexity of the investigated problem are included.

Computer simulations of hearing aid acoustical system performance

Kostek Bożena, Szwoch Grzegorz. 142nd Meeting of the Acoustical Society of America, J. Acoust. Soc. Amer., Fort Lauderdale, USA, 2001

The recent developments in the hearing aid technology enabled a number of improvements in hearing aids. This includes advanced signal processing algorithms, better speech intelligibility, miniaturization etc. One of the existing limitations is, however, the problem with providing patient-related characteristics of the acoustical system of a hearing aid. The aim of this paper is to show that using the physical modeling method it is possible to first build a model of the acoustical system of a hearing aid and then to simulate its performance. The waveguide model of the acoustical system of a hearing aid is proposed. Exemplary results of the computer simulations using such a model are presented and compared with some measurement data of existing hearing aid acoustical systems. The model proved to behave similarly to the real system. Conclusions regarding the application of such a method in the fitting process of a hearing aid are included.

Computer Modeling of Acoustical Elements of a Hearing Aid

Szwoch Grzegorz, Kostek Bożena, Czyżewski Andrzej. Archives of Acoustics, vol. 26, No. 3

In this paper, application of computer modeling methods to the process of hearing aid fitting is described. A computer model of the acoustical system of a hearing aid is presented. Exemplary results of the experiments are presented and compared with measurement data. The model proved to behave similarly to the physical system. Further improvements to the model are discussed.

Neural Computation of Direction-Of-Arrival of Sound

Czerniawski Jacek, Czyżewski Andrzej, Królikowski Rafał. 3rd WSEAS (World Scientific and Engineering Academy and Society) Int.Conf. on Neural Network and Applications (NNA '02), Interlaken, Szwjacaria, 2002

Some Rules and Methods for Creation of Surround Sound

Czyżewski Andrzej, Kornacki Artur, Odya Piotr. 21st AES Conference, Petersburg, Russia, 2002

The problem of selection of an adequate surround sound life recording and reproduction methods is still open. Alternative methods of organizing this process are discussed in the paper. Some experimental recording sessions employing the 5.1 format were made with the use of various miking techniques and the convolution-based multichannel audio processing algorithm. The results were submitted to some subjective assessments and then compared. Conclusions resulting from performed experiments are derived and discussed.

Making Surround Audio Considering Image Proximity Effect

Czyżewski Andrzej, Kostek Bożena, Odya Piotr. 112th AES Convention, Munich, Germany, 2002

The problem of influencing surround sound perception by video content was addressed employing subjective testing procedures in which experts listened to the sound with- and without video image presence and provided their answers. Results of experiments demonstrated in which cases and how video may affect the localization of virtual sound sources. The obtained data were then analyzed by means of modern techniques of intelligent data exploration and knowledge discovery allowing finding some hidden relations between semantic descriptors of subjective impressions. Finally, basing on the results of data analysis a set of rules concerning mastering of multichannel audio to accompany various types of video content were derived. Some results of this study will be presented and discussed in the paper.

Rough-Neuro Approach to Testing Influence of Visual Cues on Surround Sound Perception

Kostek Bożena. S. K. Pal, L. Polkowski, A. Skowron ed. ROUGH-NEURO COMPUTING: A WAY TO COMPUTING WITH WORDS, Springer Verlag, Series on Artificial Intelligence, 2002

Estimation of Non-Stationary Noise for Audio Enhancement by Means of Recurrent Neural Networks

Królikowski Rafał. 3rd WSEAS (World Scientific and Engineering Academy and Society) Int.Conf. on Neural Network and Applications (NNA '02), Interlaken, Szwajcaria, 2002

Soft Computing in Acoustics, Applications of Neural Networks, Fuzzy Logic and Rough Sets to Musical Acoustics, Studies in Fuzziness and Soft Computing, vol. 31

Kostek Bożena. Physica Verlag, Heidelberg, New York 1999. (ISBN3-7908-1190-4)

The book presents applications of some selected soft computing methods to acoustics and sound engineering. The aim of this research study is the implementation of soft computing methods to musical signal analysis and to the recognition of musical sounds and phrases. Accordingly, some methods based on such learning algorithms as neural networks, rough sets and fuzzy-logic were conceived, implemented and tested. Additionally, the above-mentioned methods were applied to the analysis and verification of subjective testing results. The last problem discussed within the framework of this book was the problem of fuzzy control of the classical pipe organ instrument. The obtained results show that computational intelligence and soft computing may be used for solving some vital problems in both musical and architectural acoustics.
Contents:
a. Foreword
b. Preface
c. Introduction
d. Some Selected Soft Computing Tools and Techniques: Artificial Neural Networks; Fuzzy Sets and Fuzzy Logic; Rough Sets
e. Preprocessing of Data in Acoustics: Musical Signal Representation; Musical Phrase Analyis; Acquisition of Test Results; Data Discretization
f. Automatics Classification of Musical Instrument Sounds: Uncertainty of Musical Instrument Sound Representation; Feature Vector Extraction; Statistical Properties of Musical Data; Neural Network as a Classifier of Musical Instruments; Rough Set Decision System as a Classifier of Musical Instruments
g. Automatic Recognition of Musical Phrases: Data Acquisition; Parametrization Process; Neural Network as a Classifier of Musical Phrases; Rough Set-Based Classification of Musical Phrases
h. Intelligent Processing of Test Results: Inconsistency of Subjective Assessment Results; Application of Fuzzy Logic to the Processing of Test Results; Application of Rough Sets to the Processing of Test Results; Rough-Fuzzy Method of Test Result Processing
i. Control Applications: Articulation-Related Features in the Pipe Organ Sound; Fuzzy Control of Pipe Organ
j. Conclusions