|
Lecturer(s)
|
-
Ptáček Ladislav, Ing. Ph.D.
|
|
Course content
|
Annotation: The aim of the course is to familiarise students with methods for the parameterization of acoustic signals. During this process, the original signal is transformed into feature vectors composed of selected parameters. The purpose of parameterization is to focus computational processing on relevant information. Another benefit is dimensionality reduction by eliminating redundant information and altering the temporal representation. Students will become familiar with basic parameters used primarily in the processing of human speech and bioacoustics signals. Computational methods in the time and frequency domains will be explained, including possibilities for reconstruction in the signal domain. The final part of the course will introduce sliding-window techniques and the creation of a complete feature vector for further processing. The practical part of the course will be conducted in MATLAB. 1. Characteristics and representation of acoustic signals. 2. Digitization and reconstruction of acoustic signals, coding, and modulation. 3. Basic signal operations, short-time Fourier transform, short-time DFT. 4. Time-domain processing. Characteristics, energy, autocorrelation function, zero-crossing rate. 5. Frequency-domain processing. Characteristics, spectrogram. 6. Parameterization of speech signals. Pre-emphasis, FT, DFT, band-pass filtering. 7. Cepstral analysis of speech, LPC coefficients, Mel-frequency cepstral coefficients. 8. Dynamic coefficients, energy, mean value, frequency band creation. 9. Parameterization of bioacoustics signals. Shannon entropy, ADI, AEI, peaks, normalization. 10. Sliding-window method, windowing parameter settings, filters. 11. Feature vector, statistical methods for speech recognition, GMM-UBM method for speaker recognition. 12. Delta and delta-delta coefficients, power spectral density, comparison of graphical representations. Learning outcomes: Students will acquire knowledge of methods for the parameterization of acoustic signals in both the time and frequency domains and will understand the principles of feature extraction, including cepstral analysis, LPC, and Mel-frequency cepstral coefficients. They will develop the ability to apply these methods in practice to the analysis of speech and bioacoustics signals using MATLAB and to interpret the results for recognition and classification purposes.
|
|
Learning activities and teaching methods
|
|
unspecified
|
|
Learning outcomes
|
The aim of the course is to provide students with knowledge and practical skills in acoustic signal parameterization and feature extraction. The course develops the ability to transform acoustic signals into suitable parametric representations for efficient analysis, statistical processing, recognition, and classification. Students will become familiar with methods of time- and frequency-domain analysis and with approaches used in speech and bioacoustics signal processing. Emphasis is placed on understanding the meaning of individual parameters, selecting an appropriate parameterization method with respect to the characteristics of the analysed data, and applying the methods in practice using MATLAB. The course provides a methodological foundation for further work with acoustic data within students' research activities.
|
|
Prerequisites
|
Basic knowledge of mathematics and statistics at university level and basic skills in working with digital data are expected (MSc). Basic knowledge of signal processing, acoustics, or programming is advantageous but is not a prerequisite for completing the course. Students are expected to be able to work independently with scientific literature in English. Previous experience with MATLAB is an advantage.
|
|
Assessment methods and criteria
|
unspecified
Successful completion of the course requires active participation in practical classes, completion of assigned practical tasks, and independent completion of a final project focused on the parameterization and analysis of a selected acoustic signal. Students must demonstrate the ability to select appropriate signal processing and feature extraction methods, apply them to real data, and interpret the results. The assessment includes a presentation and discussion of the final project results.
|
|
Recommended literature
|
-
] LI, F. F.; COX, T. J. Digital Signal Processing in Audio and Acoustical Engineering. 1st ed. Boca Raton: CRC Press, 2019. ISBN 978-1-0326-5223-8..
-
] SKARNITZL, R.; ŠTURM, P.; VOLÍN, J. Zvuková báze řečové komunikace: Fonetický a fonologický popis řeči. Praha: Karolinum, 2016. ISBN 978-80-246-3300-8. ] HILL, P. Audio and Speech Processing with MATLAB. 1st ed. Boca Raton: CRC Press, 2019. ISBN 978-1-4987-6274-8..
-
UHLÍŘ, J.; SOVKA, P.; POLLÁK, P.; HANŽL, V.; ČMEJLA, R. Technologie hlasových komunikací. Praha: Nakladatelství ČVUT, 2007..
-
VESELÝ, V.; RAJMIC, P.; MOKRÝ, O. Moderní funkcionální analýza s aplikacemi ve zpracování signálů. Brno: VUTIUM, 2024. ISBN 978-80-214-6138-3..
|