Approach for spectrogram analysis in detection of selected pronunciation pathologies
An attempt to automatise selected pronunciation pathology detection in preschool children is described in this paper. Consonant [Z] in various phonetic surroundings is taken into consideration as eventual sigmatism indicator. The analysis involves spectrogram analysis in terms of image processing methods used for feature extraction and classification. Five dedicated features are defined and extracted, i.a., from a frequency sub-band of [1500, 6500] Hz. Binary classification using support vector machine enables pathology detection. The system performance is evaluated using sensitivity, specificity, and accuracy metrics in two cross-validation experiments over a database of 140 speech recordings with 50 normative and 90 pathological cases. Repeatable efficiency metrics at a ca. 85% level confirm the method capabilities and encourage to develop the system for the speech diagnosis support.
Keywordsspeech pathology spectrogram analysis image processing
Unable to display preview. Download preview PDF.
This research was supported by the Polish Ministry of Science and Silesian University of Technology statutory financial support No. BK-200/RIB1/2017.
- 1.R. Tadeusiewicz. Sygnał mowy. Wydawnictwa Komunikacji i Łączności, Warszawa, 1988 (in Polish).Google Scholar
- 2.A. Kaczmarek. Analiza sygnału mowy w foniatrii. Oddział Gdański PTETiS, 2006 (in Polish).Google Scholar
- 3.Z. Miodonska, M. D. Bugdol, and M. Krecichwost. Dynamic time warping in phoneme modeling for fast pronunciation error detection. Computers in Biology and Medicine, 69:277–285, 2016.Google Scholar
- 4.M. Krecichwost, Z. Miodonska, J. Trzaskalik, J. Pyttel, and D. Spinczyk. Acoustic Mask for Air Flow Distribution Analysis in Speech Therapy. In Information Technologies in Medicine, ITIB 2016, vol. 1, volume 471 of Advances in Intelligent Systems and Computing, pages 377–387, 2016.Google Scholar
- 5.T. Lampert and S. O’Keefe. A survey of spectrogram track detection algorithms. Department of Computer Science, University of York, 2008.Google Scholar
- 6.T. Lampert and S. O’Keefe. On the detection of tracks in spectrogram images. Department of Computer Science, University of York, 2012.Google Scholar
- 7.P.J. Durka. Między czasem a częstoscią: elementy współczesnej analizy sygnałów. 1999 (in Polish).Google Scholar
- 8.B. Pinkowski. Principal component analysis of speech spectrogram images. Computer Science Department, Western Michigan University, 1996.Google Scholar
- 9.R. Gonzalez and R.Woods. Digital Image Processing. Prentice Hall, Upper Saddle River, NJ, 2002.Google Scholar
- 10.O. Farooq and S. Datta. Mel Filter-Like Admissible Wavelet Packet Structure for Speech Recognition. IEEE Signal Processing Letters, vol. 8, No. 7, July 2001.Google Scholar
- 11.C.J. Long and S. Datta. Wavelet based feature extraction for phoneme recognition. Spoken Language, 1996. ICSLP 96. Proceedings., Fourth International Conference, 1996.Google Scholar
- 12.L. Deng, G. Hinton, and B. Kingsbury. New types of deep neural network learning for speech recognition and related applications: an overview. In 2013 IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, May 2013.Google Scholar
- 13.A. Waibel, T. Hanazawa, G. Hinton, K. Shikano, and K. J. Lang. Phoneme recognition using time-delay neural networks. IEEE Transactions on Acoustics, Speech, and Signal Processing, 37(3):328–339, Mar 1989.Google Scholar
- 14.Z. Miodonska, M. Krecichwost, and A. Szymanska. Computer-Aided Evaluation of Sibilants in Preschool Children Sigmatism Diagnosis. In Information Technologies in Medicine, ITIB 2016, vol. 1, volume 471 of Advances in Intelligent Systems and Computing, pages 367–376, 2016.Google Scholar
- 15.C. Cortes and V. Vapnik. Support-vector networks. Machine Learning, 20(3):273–297, 1995.Google Scholar
- 16.S. Arlot and A. Celisse. A survey of cross-validation procedures for model selection. Statistics Surveys, 4:40–79, 2010.Google Scholar