04-801-I4   Speech Signal Processing

Course description

Voice-enabled applications such as information access, speech translation, and voice biometrics are becoming popular due to the penetration of mobile phones and the ease of communication using speech mode. The success in developing these applications depends on a good understanding of the speech signal and the methods to extract information about the speech production mechanism. This course is aimed at giving basics of signal processing and methods for processing speech signals. The course will be a self-contained course, as all the necessary background of digital signal processing and random processes will also be covered. There will also be a discussion on how speech recognition and other applications are achieved using statistical and neural network models.

Learning objectives

This course covers the basics of speech signal processing methods for extracting features of speech production mechanisms that are needed for developing voice-based applications.

Content details

• Background to speech signal processing (1 lecture)
• Speech production mechanism and nature of the speech signal (2 lectures)
• Basics of digital signal processing and equivalent representations of signals and systems (3 lectures)
• Introduction to random processes (1 lecture)
• Speech signal processing methods: Short-time spectrum analysis and linear prediction analysis (4 lectures)
• Hidden Markov Models: An introduction to speech recognition (3 lectures)

Prerequisites

The course is mostly self-contained, but some background in linear algebra and probability theory will be useful. Basic mathematics covered in engineering programs is assumed.

Faculty

B. Yegnanarayana