Speech compression with wavelet packets

Coefficient decomposition in an orthogonal basis performs a decorrelation that can be applied to reduce the information needed to transmit the highly correlated samples that characterize signals in the time domain. Moreover, by changing the orthonormal basis, the number of coefficients that support this information can be minimized. Searching over a library of orthonormal basis means performing a wavelet packet decomposition. Making use of this property, this paper present a wavelet packet approach to the signal compression problem. After a description of the wavelet packet application, experimental results are obtained in order to establish the performance of wavelets in transform based coders