Efficient Coding Methods for a Perceptual Speech Coder

Project Type: Master/Diploma Thesis
Student: Stocker Markus


 The aim of this work is to find an efficient data compression algorithm for the perceptual speech coder proposed by G. Kubin and W.B. Kleijn (ICASSP 1999). The compression algorithm should enable to transmit the signal representation through a channel with a low bit rate. First, a brief introduction on information theory and the explanation of its most important terms is given. After that the perceptual speech coder is introduced and explored from the angle of data reduction. The next step is the application of current lossless data compression algorithms to the representation of the original signal obtained by the speech coder. Finally, the studies of vector quantization and of the JBIG-standard show possibilities and problems of the development of a coding algorithm which optimal fits to the given data structure. Starting from an original speech signal with 16 kHz sampling rate and 16 bit amplitude resolution, we have a data rate of 256 kbit/sec. With the application of the different coding algorithms to the representation of the original signal the data rate can be reduced by a factor of 2.4 to a data rate of 105 kbit/sec.