Digital Speech Processing
Fundamental knowledge on digital speech processing, speech modeling and analysis, recognition and synthesis. Experience in digital speech processing applications in communication and multimedia systems.
- recognize the significance of digital speech processing and its applications
- describe speech production mechanism and corresponding physical models
- list various methods for modeling of speech signal in continuous and discrete time domain
- apply linear prediction methods for modeling of speech signal
- employ homomorphic speech processing for estimation of excitation and vocal tract model
- develop simple algorithms for speech processing using Matlab
- analyze quantization effects of model coefficients on its accuracy
- apply methods for recognition of vowels and speaker identity
Forms of Teaching
Lectures are organized in two terms. First term consists of 7 weeks of lectures and midterm exam. The second term consists of another 6 weeks of lectures and final exam. Weekly workload for lectures is 2 hours for total of 15 weeks in semester.Exams
Course can be passed through continuous assesment based on midterm and final exams. Final exam consists of written part and oral exam. Students have another possibility to pass the course through end of semester exam, that also has a written and oral part.Laboratory Work
During the semester, two laboratory exercises are organized in accordance to week-by-week plan. These exercises are used to prepare students for individual work.Consultations
Consultations are organized in lecture weeks just after the lectures.Programming Exercises
Total course workload related to student individual work amounts to 90 hours, which students use for Program exercises and preparation for exams. Homework for each of the two semester terms is the Report of individual work on Program exercises. This Report also includes the report of Laboratory exercises. For individual work, students have to examine corresponding chapters in Course-book and Lecture notes which are cited in the week-by-week plan, perform the required exercises and prepare the report for each chapter.
|Type||Threshold||Percent of Grade||Threshold||Percent of Grade|
|Laboratory Exercises||0 %||10 %||0 %||10 %|
|Homeworks||0 %||20 %||0 %||20 %|
|Mid Term Exam: Written||0 %||30 %||0 %|
|Final Exam: Written||0 %||30 %|
|Final Exam: Oral||10 %|
|Exam: Written||50 %||60 %|
|Exam: Oral||10 %|
Assessment of Laboratory exercises and Homework (Program exercises) is performed commonly based on submitted Reports of individual work for the first and second term. Students can approach the oral part of the final exam only if they have at least 50% of total points from midterm exam and written part of the final exam.
Week by Week Schedule
- Lectures: (L): Introduction to digital speech processing and its applications, Automatic speech, speaker and language recognition, Basic principles of speech synthesis, Text-to-Speech, Computer dialog systems with applications in virtual reality; Individual work (I): Chap: Survey of digital speech processing applications, Chap: Fundamentals of speech production, Chap: Phonetics and Linguistics.
- Lectures (L): Fundamentals of speech production, Physical model of production; Individual work (I): Cahp 1: Recording of speech signals using sound cards.
- Lectures (L): Acoustic model of vocal tract, Excitation signal of the vocal tract; Individual work (I): Chap 2: Analysis of speech signals in time domain.
- Laboratory (LB): Chap 4: Analysis of speech formant structure; Individual work (I): Chap 3: Spectral analysis of speech signals and spectrograms.
- Lectures (L): Connected tube model of the vocal tract, Time discrete vocal tract model; Individual work (I): Chap 5: Automatic classification of vowels based on their format structure.
- Lectures (L): Linear prediction and its application for speech modeling; Individual work (I): Chap 6: Automatic speaker classification based on formant structure.
- Lectures (L): Autocorrelation method for LPC model estimation; Individual work (I): Chap 7: Linear prediction methods.
- Midterm exam
- Laboratory (LB): Chap 8: Autocorrelation method for speech predictor estimation; Individual work (I): Chap 9: Levinson-Durbin algorithm; prediction gain analysis.
- Lectures (L): Covariance method for LPC model estimation, Parametric representations for short-time speech spectral envelope modeling; Individual work (I): Chap 10: Covariance method for speech predictor estimation.
- Lectures (L): Homomorphic speech processing; Individual work (I): Chap 11: Quantization effects of LPC predictor coefficients.
- Lectures (L): Applications of homomorphic processing on speech signal; Individual work (I): Chap 12: Homomorphic analysis of speech signal.
- Lectures (L): Introduction to automatic speech recognition (ASR), Speech analysis for ASR; Individual work (I): Chap 13: Voicing and pitch estimation.
- Lectures (L): Feature vectors; Statistical models and classification methods for ASR; Individual work (I): Chap 14: Example of the Vocoder.
- Final exam