Speech Compression Using Discrete Wavelet Transform

• Main Author: M.Ali Najih, Abdulmawla
• Format: Thesis
• Language: English; English
• Published: 2003
• Subjects: Wavelets (Mathematics); Compression
• Online Access: http://psasir.upm.edu.my/id/eprint/12219/1/FK_2003_57.pdf; http://psasir.upm.edu.my/id/eprint/12219/
• Institution: Universiti Putra Malaysia

Speech compression is an area of digital processing that is focusing on reducing bit rate of the speech signal for transmission or storage without significant loss of quality. Wavelet transform has been recently proposed for signal analysis. Speech signal compression using wavelet transform is given a considerable attention in this thesis. Speech coding is a lossy scheme and is implemented here to compress onedimensional speech signal. Basically, this scheme consists of four operations which are the transform, threshold techniques (by level and global threshold), quantization, and entropy encoding operations. The reconstruction of the compressed signal as well as the detailed steps needed are discussed.The performance of wavelet compression is compared against linear Productive Coding and Global System for Mobile Communication (GSM) algorithms using SNR, PSNR, NRMSE and compression ratio. Software simulating the lossy compression scheme is developed using Matlab 6. This software provides the basic speech analysis as well as the compression and decompression operations. The results obtained show reasonably high compression ratio and good signal quality.