Speech Compression with Wavelet Transform and Huffman Coding
Ranjushree Pal · 2021
The speech signals as it is generated carry lot of information which is redundant and has to be discarded before efficient storage and transmission. Thus, an efficient technique of speech compression without loss of much information giving less distortion in the reconstructed waveform and also retaining the perceptual qualities of speech is very much needed. This paper deals with the recording of female speech signals in Praat software. Wavelet Transform is carried out on these signals which removes lot of redundant information followed by Huffman coding which compresses the signals further. The average compression ratio obtained is three times higher than that obtained with only Wavelet Transform. The reconstructed signals have very less distortion (10-4) while retaining the maximum percentage of signal energy and perceptual qualities of speech. The detailed results are summarized in the concluding section.