Vector quantization and its application to speech coding

• Main Author: Chen, Changqian.
• Other Authors: Koh, Soo Ngee
• Format: Theses and Dissertations
• Language: English
• Published: 2008
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence; DRNTU::Engineering::Computer science and engineering::Data::Coding and information theory
• Online Access: http://hdl.handle.net/10356/13141
• Institution: Nanyang Technological University

This thesis is devoted to the investigation of vector quantization for speech coding. The soundness of the code-excited linear prediction (CELP) and multi-band excitation (MBE) speech production models has been verified by the fact that the speech coders based on these models are able to synthesize high quality speech signals. However, such coders do not perform satisfactorily at bit rates lower than certain known thresholds. This opens up the possibility for the bit rates necessary for satisfactory speech representation to be reduced further, by describing more concisely the LPC parameters of the CELP model and the spectral magnitudes in the MBE model. That is, efficient representation of model parameters is a means to achieving or improving low bit rate speech coding, although efforts to enhance the model itself may also lead to some improvements.