Institute of Formal and Applied Linguistics
Faculty of Mathematics and Physics
Charles University in Prague

is pleased to invite you to the lecture


"A Novel Approach to Speech Coding After Time Scale Modification"

by H. Gokhan Ilk, Ph.D.,
Ankara University, Faculty of Engineering, Electronics Engineering Department


The lecture will be held on Monday, September 17, 2007, at 2 p.m.
at Faculty of Mathematics and Physics,
Malostranske namesti 25, Praha 1,
4th floor, room S1 (Nr. 428).

--------
Abstract

Speech is the most conventional means of communication amongst human
beings. Besides its social aspects, it is perhaps the most important
random statistical signal that has attracted enormous attention in order
to develop efficient algorithms. One set of these algorithms is the
efficient coding of speech.

Although there are plenty of speech coding algorithms developed and it is
impossible to list them in a seminar, it is possible to say that Linear
Prediction has been successfully applied to speech coding and many coders
are based on this technique. Although Linear Prediction is not only
applicable to speech, the main reason for the success of Linear Prediction
in speech is that the autoregressive model combines good prediction
capabilities in the time domain with an accurate description of the speech
spectral envelope.

In this seminar conventional LPC (Linear Predictive Coding) based speech
coding algorithms are introduced by describing the linear prediction
initially. Innovations representation, auto correlation and covariance
methods are described in detail. This will form the basic understanding of
LPC analyses and synthesis filters and conclude the current state of art
of speech coding algorithms.

The second part of the seminar will describe a novel approach designed by
the author that uses conventional speech coding algorithms after time
scale modification. WSOLA (Waveform Similarity Overlap and Add) has been
used as the time scale modification algorithm. The signal is adaptively
compressed before coding with a time-dependent compression rate in order
to use the available bandwidth most efficiently. This approach is
particularly useful in VoIP applications in dynamic networks because the
load may change abruptly and it is not symmetric at each direction. The
author hopes that one day all vocoders (voice coders) will be designed in
this way.

--------
Short CV of Dr. H. Gokhan Ilk

Dr. Gokhan Ilk received the B.Sc. in 1993 from Ankara University,
Department of Electronics Engineering and the M.Sc. and the PhD degrees
from the University of Manchester Institute of Science and Technology
(UMIST) U.K. in 1994 and 1997, respectively. He is currently a member of
staff at the Ankara University, Electronics Engineering Department, where
he teaches courses and conducts research on Digital Signal Processing,
Speech Processing and Coding, as well as several programming languages and
techniques.