Speaker: Dr. Ronald Schafer

Title: The Cepstrum: Its History and Role in Speech Processing

Date: March 19, 2004

Time: 3:00 pm

Location: GCATT Room 325

Abstract:
The term cepstrum was coined by Bogert, Healy and Tukey in a paper in an edited book published in 1965. The term was designed to reflect the change of perspective that results when one transforms a signal by taking the inverse Fourier transform of the logarithm of its power spectrum. Bogert et al introduced the cepstrum in the context of detection of echos. At about the same time, Oppenheim (Ph.D. thesis, 1964) proposed a new class of systems called homomorphic (or generalized linear) systems. A subclass of these systems (homomorphic systems for convolution) can be represented in terms of cepstrum-like operations. Schafer (Ph.D. thesis, 1968) and Oppenheim generalized the cepstrum to the complex cepstrum, defined as the inverse Fourier transform of the complex logarithm of the Fourier transform of the signal, and showed that the complex cepstrum could be used for separating (or filtering) the components of a convolution of two or more signals.
Early on, it became obvious that the cepstrum of a speech signal had many interesting and useful properties. This led many speech researchers to apply the cepstrum in areas such as pitch detection, formant estimation, speaker verification, and speech recognition.
This paper will begin with a brief introduction to the history of the cepstrum concept. Most of the talk will focus on the many ways that the cepstrum has been used in speech processing during the past 25 years. The talk will conclude with some new ideas that promise to stimulate renewed interest in the cepstrum and its applications to speech signal processing problems.

Biography:
http://users.ece.gatech.edu/~rws/

Slides:

Speaker:	Dr. Ronald Schafer
Title:	The Cepstrum: Its History and Role in Speech Processing
Date:	March 19, 2004
Time:	3:00 pm
Location:	GCATT Room 325
Abstract:	The term cepstrum was coined by Bogert, Healy and Tukey in a paper in an edited book published in 1965. The term was designed to reflect the change of perspective that results when one transforms a signal by taking the inverse Fourier transform of the logarithm of its power spectrum. Bogert et al introduced the cepstrum in the context of detection of echos. At about the same time, Oppenheim (Ph.D. thesis, 1964) proposed a new class of systems called homomorphic (or generalized linear) systems. A subclass of these systems (homomorphic systems for convolution) can be represented in terms of cepstrum-like operations. Schafer (Ph.D. thesis, 1968) and Oppenheim generalized the cepstrum to the complex cepstrum, defined as the inverse Fourier transform of the complex logarithm of the Fourier transform of the signal, and showed that the complex cepstrum could be used for separating (or filtering) the components of a convolution of two or more signals. Early on, it became obvious that the cepstrum of a speech signal had many interesting and useful properties. This led many speech researchers to apply the cepstrum in areas such as pitch detection, formant estimation, speaker verification, and speech recognition. This paper will begin with a brief introduction to the history of the cepstrum concept. Most of the talk will focus on the many ways that the cepstrum has been used in speech processing during the past 25 years. The talk will conclude with some new ideas that promise to stimulate renewed interest in the cepstrum and its applications to speech signal processing problems.
Biography:	http://users.ece.gatech.edu/~rws/
Slides: