Speech compression using the discrete wavelet transform

• Main Author: Lim, Russel Lloyd C.
• Format: text
• Published: Animo Repository 2003
• Subjects: Data compression (Computer science); Wavelets (Mathematics); Numerical Analysis and Scientific Computing
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/7725
• Institution: De La Salle University

Wavelet analysis is a relatively new technology. Wavelets are waveforms that occur in a very short duration that has a mean value of zero. Wavelet transform can represent non-stationary signals more effectively than Fourier transform since it retains both the time and frequency aspect of the signal. This thesis applies wavelet analysis to speech compression. A mother or basis wavelet is first chosen for the compression. The signal is then decomposed to a set of scaled and translated versions of the mother wavelet. The resulting wavelet coefficients that are insignificant or close to zero are truncated achieving signal compression. Additional compression is realized by encoding of the signal. Analysis of the compression process was performed by comparing the compressed-decompressed signal against the original. This was conducted to determine the effect of the choice of mother wavelet on the speech compression. The results however showed that regardless of bases wavelet used the compression ratio is relatively close to one another. In terms of signal quality, coiflet of order 5 has been seen to be the best basis wavelet. This is taken from the analysis of the Signal to Noise Ratio (SNR) value and its Mean Opinion Score (MOS) rating. Moreover, a comparative analysis of current technologies against these results was performed. The system developed proved to be very efficient in terms of compressing speech files since the system was able to compress better than the current technology, Motion Pictures Expert Group -1 Layer 3 (MP3). The system was able to reduce a specific speech file to 57.41 % while MP3 with 112 kbps bit rate rated at 69.27% against the original signal. Another speech signal was reduced to 58.31 % while the MP3 rated at 70.55%, against the original speech signal. The drawback however is that signal fidelity produced by the system was unable to match the signal quality produced by the current technology, MP3.