01-09-2012, 10:18 AM
Implementation of Speech Compression using Wavelet Transforms on TMS320DM642 DSP Processor
1Implementation of Speech.doc (Size: 23 KB / Downloads: 31)
ABSTRACT
The project “Implementation of Speech Compression using Wavelet Transforms on TMS320DM642 DSP Processor” aims at compressing a given speech signal using wavelet transforms and implementing it on heterogeneous platform, which uses TMS320DM642 DSP Processor. There are many compression techniques being used to compress speech like Fourier transforms, short-time Fourier transforms. But every compression technique has its own drawback. Compression of speech using Wavelet transforms gives solution to all the problems of other compression techniques. There are many wavelet families available such as Haar, Daubechies, Bi-orthogonal wavelets. Db4 wavelets are used to compress the speech signal as it provides higher SNR ratio.
In this project the DWT filters are implemented in real time using TMS320DM642 DSP processor. This TMS320DM642 heterogeneous EVM kit is used as it provides higher computational speed. To this heterogeneous platform, a real time speech signal i.e, a human voice is given as input to the processor through the microphone and the processed output is observed through the speakers.
From the simulations of the MATLAB and CCS, the compressed and reconstructed signals are observed and a high SNR ratio is obtained. The SNR ratios for different signals are observed to prove that signal is perfectly reconstructed. Also, the speech signal given out by the speakers can be heard clearly through the speaker, which proves that the compression and reconstruction of the speech signal is performed perfectly using DWT filters.