Prerequisites by Topic:

Discrete time signals and systems
Z-transform, discrete Fourier transform, FFT
Digital filters, Probability and random process

Speech models: acoustic-tube, Source-filter, Time and frequency domain properties, Linear prediction analysis of speech, Speech coding: DPCM, ADPCM, sub-band, VQ, etc. Speech synthesis and recognition, Speech processing hardware. (3 credits)

To familiarize the student with principles of digital speech signal processing and its applications.

• Discrete-Time Processing of Speech Signals
  J. R. Deller, Jr., J. H. L. Hansen, and J. G. Proakis, IEEE Press, 1999. (ISBN# 0780353862)

An extensive list of references (books and journals) will be provided in the class to introduce recent advancements in the field. Course handouts using web based modules will be supplemented.

• Office Phone: (813) 974-4769; Office Location: ENB 368
• Fax: (813) 974-5250; Lab Phone: (813) 974-3510
• E-mail: sankar@eng.usf.edu

Office Hours: MW 11:00-1:00 pm but mostly open door policy
You can also contact me by email/voice mail any time or by appointment.

Course Homepages:

• USF Web Portal - http://my.usf.edu (access myUSF Online and Blackboard)
• Mirror site - http://www.eng.usf.edu/~sankar/course/speech.html

Topics to be covered:

• Introduction to speech signal processing
  Reviews of signal processing and mathematics for speech processing

• Speech models: speech production and acoustic-phonetics
  Digital filter models

• Speech analysis: time and frequency domain representations
  Short-time energy, autocorrelation, and power spectra; Linear prediction and Cepstral representations

• Linear predictive coding: Levinson-Durbin and Schur algorithms
  Lattice filters, Applications of LPC for speech processing

• Speech coding: Waveform Coders: DPCM, ADPCM, and SBC
  Parametric Coders, Hybrid Coders (CELP and Low-bit rate coders), VBR Coding, Vector quantization (VQ) etc.

• Speech recognition: Pattern recognition model, Algorithms for recognition:
  endpoint detection, feature extraction, DTW, and decision measures
  Network models for recognition

• Speech applications: robust system design, speech synthesizers, text-to-speech systems, etc

Homework - 10%
Computer Assignments (MATLAB) - 40%
Project (Technical Paper Analysis/Software) - 25%
Exam - 25%

Academic Policies:

Homework Policy: Homework Exercises will be assigned in the class. Everyone is recommended to do the homework earnestly since it will be a good preparation for the exam.

Exam Policy: All exams are closed books and notes. One page reference sheet for formulas and definitions is allowed but NO homework or any other worked out examples. There will be NO MAKE-UP for a missed exam without prior approval from the instructor (with sufficient advance notice given) except in the case of a documented medical emergency. All students must take the exams during regularly scheduled class or exam times either on campus or with an approved proctor. Any deviation from this policy MUST be pre-approved by the instructor in writing.

Attendance Policy: Class attendance is NOT mandatory but the probability of success (i.e., securing a good grade) in the course is very low for a student with sporadic attendance record. Students who anticipate the necessity of being absent from class due to the observation of a major religious observance must provide notice of the date(s) to the instructor, in writing, by the second class meeting.

Academic Dishonesty Policy: Students are reminded that the Department policy dictates that any form of cheating or plagiarism will result in a "FF" grade in the course and further suspension or expulsion from the University. Receiving or providing help on exams and not submitting individual work on assignments and project are forms of cheating; Submissions that are "identical" in any way are clear evidence of cheating. Copying materials from textbooks and papers without properly referencing them or not giving due credit are forms of plagiarism. It is the student's responsibility to review and understand USF and EE Department policies and procedures on Academic Conduct, Dishonesty, and Disruption.