The compression of musical instrument signals by wavelet transform

The performance of the wavelet transform in the analysis, synthesis and compression of the audio signals of some musical instruments and of speech signals is investigated. Musical samples are taken from flute, bass guitar, clarinet and cello. In the analysis and the synthesis, Mallat's algorithm is used to calculate the wavelet transform. The process is repeated for all wavelet functions that can be calculated. The performances are evaluated by SNR (signal-to-noise ratio) values as an objective criterion and an MOS (mean opinion score) experiment has been carried out for a subjective criterion. The results are given and they show that the compression rate of the wavelet transform is very high, even though the system of human perception and any lossless compression algorithm have not been taken into account.