Selected publications
See also:
Old publications (abstracts only)
Selected publications
Pitch Detection Enhancement Employing Music Prediction
Marek Szczerba, Andrzej Czyżewski
Journal of Intelligent Information Systems, Vol. 24, No. 2-3. (March 2005), pp. 223-251.
"Computing with words" Concept Applied to Musical Information Retrieval
Bożena Kostek
Electronic Notes in Theoretical Computer Science, vol. 82, No. 4, 2003.
The objective of the paper is to provide cognitive-based mechanisms underlying processing of musical instrument sounds. The system proposed by the author based on the rough set method and on fuzzy logic provides knowledge on how humans internally represent such notions as quality and timbre and therefore it allows for the human-like automatic processing of musical data. Therefore "Computing with words" concept can be used in musical information retrieval domain by offering better processing of subjective descriptors of musical instrument sounds and enabling the analysis of data that would result in extraction of semantic information related to musical instrument sounds. This paper shows first a review of developments in the domain of timbre mapping and classification. A decision table is built of semantic descriptors of musical instrument sounds, then rules extracted by the rough set method and the processing of musical timbre based on fuzzy logic is shown. An example of rough-fuzzy processing is given and conclusions are derived.
Representing Musical Instrument Sounds for Their Automatic Classification
Bożena Kostek, Andrzej Czyżewski
J. Audio Eng. Soc., vol. 49, No. 9, 768-785, 2001.
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 standardization process are included.
Rough-Neuro Approach to Testing the Influence of Visual Cues on Surround Sound Perception
Bożena Kostek
Chapter in ROUGH-NEURO COMPUTING: A WAY TO COMPUTING WITH WORDS (S. K. Pal, L. Polkowski, A. Skowron eds.), Springer Verlag, Series on Artificial Intelligence, pp. 555 - 572, 2004.
This chapter aims at revealing in which way and how surround sound interferes or is associated with visual context. Such parameters as distance, angle, or level of sound source were tested with and without a video image on the screen. For that purpose, a subjective testing was applied. Processing of the results obtained was done by employing genetic algorithms and combined neural network and rough set systems. The main task of the experiments was the application of modular neural networks to quantize surround sound parameter values. A rough set algorithm was used to make decisions showing the influence of visual cues on the perception of surround sound.
High Accuracy and Octave Error Immune Pitch Detection Algorithms
Marek Dziubiński, Bożena Kostek
Archives of Acoustics, vol. 29, No. 1, 1-21, 2004.
The aim of this paper is to present a method improving pitch estimation accuracy, showing high performance for both synthetic harmonic signals and musical instrument sounds. This method employs an Artificial Neural Network of a feed-forward type. In addition, octave error optimized pitch detection algorithm, based on spectral analysis is introduced. The proposed algorithm is very effective for signals with strong harmonic, as well as nearly sinusoidal contents. Experiments were performed on a variety of musical instrument sounds and sample results exemplifying main issues of both engineered algorithms are shown.
Employing Fuzzy Logic and Noisy Speech for Automatic Fitting of Hearing Aid
Bożena Kostek, Andrzej Czyżewski
142 Meeting of the Acoustical Society of America, No. 5, vol. 110, preprint 2pPP10, pp. 2680, Fort Lauderdale, USA, 3.12.2001 - 7.12.2001.
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 anew 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.
Principles and acoustical foundations of the computer-based hearing screening method
Henryk Skarżyński, Andrzej Czyżewski, Bożena Kostek
144th Meeting of the Acoustical Society of America (First Pan-American/Iberian Meeting on Acoustics), J. Acoust. Soc. Am., No. 5, vol. 112, Cancun, Mexico, 2.12.2002 - 7.12.2002.
The hearing impairment is one of the fastest growing diseases in modern societies. Therefore it is very important to organize screening tests allowing to find people suffering from this kind of impairment. The computer-based system was designed to conduct hearing screening, mainly in children and youth. The test uses automatic questionnaire analysis, audiometric tone test procedure and testing speech intelligibility in noise. The starting point of the test, is an automatic interview with the individual to be tested. Based on the interview, the electronic questionnaire is filled out. After the questionnaire has been filled out and the specially conceived three tone audiometric test is completed, it might be selected the mode of the speech-in-noise based test as appropriate for the specific age. When all the testing is completed, the system "I CAN HEAR..." automatically analyzes the results for every person examined. Based on the number of wrong answers those who may have hearing problems are referred to co-operating medical consulting centers. In the paper foundations and principles of the hearing tests are discussed and results of testing of more than 200.000 children with this method are demonstrated.
Diagnostic system for speech articulation and speech understanding
Andrzej Czyżewski, Bożena Kostek, Henryk Skarżyński
144th Meeting of the Acoustical Society of America (First Pan-American/Iberian Meeting on Acoustics), J. Acoust. Soc. Am., No. 5, vol. 112, Cancun, Mexico, 2.12.2002 - 7.12.2002.
With the increase in access to multimedia computers, speech training can be made available to patients with no continuous assistance required from speech therapists. Another function the system can easily perform is screening testing of speech fluency providing directed information to patients who have various speech disorders and problems with understanding speech. The idea underlying the proposed system is a programmed speech therapy training algorithm consisting of diagnostic tools and rehabilitation devices connected with it. The first function the system has to perform is data acquisition where information about the patient's medical history is collected. This is done through electronic questionnaires. The next function is analysis of the speech signal articulated by the patient when prompted by the computer followed by some multimedia tests carried out in order to assess the subject's ability to understand speech. Next, the results of the electronic questionnaire, the patient's voice and patient's reactions are automatically analyzed. Based on that the system automatically diagnoses possible speech disorders and how strong they are. A large number of schoolchildren was tested with this method. In the paper foundations of applied speech testing method and obtained results will be demonstrated.
Prediction of the Reverberation Time in Rectangular Rooms with Non-Uniformly Distributed Sound Absorption
R. Neubauer, Bożena Kostek
Archives of Acoustics, vol. 26, No. 3, 183 - 201, 2001.
The aim of this paper is first to review the best known reverberation time formulae and then to show that the reverberation time cannot be thereby predicted accurately in cases mostly encountered in practice, where the sound field is not diffuse. Introducing a correction to Fitzroy's formula allows better prediction of 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. In addition, the results obtained by measuring reverberation conditions in situ and those predicted for the same enclosure are compared and conclusions drawn.
Intelligent Processing of Stuttered Speech
Andrzej Czyżewski, Andrzej Kaczmarek, Bożena Kostek
Journal of Intelligent Information Systems, vol. 21, No. 2, pp. 143 - 171, 2003.
The process of counting stuttering events could be carried out more objectively through the automatic detection of stop-gaps, syllable repetitions and vowel prolongations. The alternative would be based on the subjective evaluations of speech fluency and may be dependent on a subjective evaluation method. Meanwhile, the automatic detection of intervocalic intervals, stop-gaps, voice onset time and vowel durations may depend on the speaker and the rules derived for a single speaker might be unreliable when trying to consider them as universal ones. This implies that learning algorithms having strong generalization capabilities could be applied to solve the problem. Nevertheless, such a system requires vectors of parameters, which characterize the distinctive features in a subject's speech patterns. In addition, an appropriate selection of the parameters and feature vectors while learning may augment the performance of an automatic detection system. The paper reports on automatic recognition of stuttered speech in normal and frequency altered feedback speech. It presents several methods of analyzing stuttered speech and describes attempts to establish those parameters that represent stuttering event. It also reports results of some experiments on automatic detection of speech disorder events that were based on both rough sets and artificial neural networks.
Some Rules and Methods for Creation of Surround Sound
Andrzej Czyżewski, Artur Kornacki, Piotr Odya
21st AES Conference, Petersburg, Russia, 1.6.2002 - 3.6.2002.
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.
Intelligent Acquisition of Audio Signal Employing Neural Network and Rough Set Algorithms
Andrzej Czyżewski
Chapter 20 in ROUGH-NEURO COMPUTING: A WAY TO COMPUTING WITH WORDS (S. K. Pal, L. Polkowski, A. Skowron eds.), Springer Verlag, Series on Artificial Intelligence, pp. 521 - 541, 2004.
The algorithms stemming from the neuro-rough computing approach were applied to digital acquisition of audio signals with regard to automatic localization of sound sources with the presence of noise and parasite echo. The application of neural networks to the automatic detection of sound arrival direction was tested first, then it was followed by some experiments employing rough sets and finally the neuro-rough approach to this problem solving was examined. The output of each tested algorithm was supposed to provide information about the direction of arriving sound. In the case of the neuro-rough algorithm the result of its action can be also available in the form of words defining the direction of arriving sound. Some details of the engineered systems and results of their experimental verification are compared and discussed.
Pitch Estimation Enhancement Employing Neural Network-Based Music Prediction
Andrzej Czyżewski, Marek Szczerba
Proc. IASTED Intern. Conference, Artificial Intelligence and Soft Computing, 413 - 418, 17.7.2002- 19.7.2002, Banff, Canada, 2002.
In this paper a new method for pitch estimation enhancement was presented. Pitch estimation methods are widely used for extracting musical data from digital signal. A brief review of these methods is included in the paper. However, since processed signal may contain noise and distortions, the estimation results can be erroneous. The proposed method was developed in order to override disadvantages of standard pitch estimation algorithms. The new -approach is based on both pitch estimation in terms of signal processing and pitch prediction based on musical knowledge modeling. First, signal is partitioned into segments roughly analogous to consecutive notes. Thereafter, for each segment an autocorrelation function is calculated. Autocorrelation function values are then altered using pitch predictor output. A music predictor based on artificial neural networks was introduced for this task. The description of the proposed pitch estimation enhancement method is included and some details concerning music prediction are discussed in the paper.
Neuro-Rough Control of Masking Thresholds for Audio Signal Enhancement
Andrzej Czyżewski, Rafał Królikowski
Neurocomputing. An International Journal, vol. 36, No. 1-4, pp. 5-27, 2001.
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.
Automatic Identification of Sound Source Position Employing Neural Networks and Rough Sets
Andrzej Czyżewski
Pattern Recognition Letters, vol. 24, pp. 921-933, 2003.
Methods for the identification of direction of the incoming acoustical signal in the presence of noise and reverberation were investigated. Since the problem is a non-deterministic one, thus applications of two learning algorithms, namely neural networks and rough sets were developed to solve it. Consequently, two sets of parameters were formulated in order to discern target source from unwanted sound source position and then processed by learning algorithms. The applied feature extraction methods are discussed, training processes are described and obtained sound source localizing results are demonstrated and compared.
Computer Modeling of Acoustical Elements of a Hearing Aid
Grzegorz Szwoch, Bożena Kostek, Andrzej Czyżewski
Archives of Acoustics, vol. 26, No. 3, 203-213, 2001.
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.