Speech Coding with Wavelet Packet Excitation Signal Compression*

Digital speech coding is important problem in the area of network speech applications where there is a necessity of transmission of big amount of data through limited bandwidth channels. Investigations in this field aim at finding new methods for signal compression and compatibility with the network transfer protocols. Digital speech coding involves sampling and amplitude quantization of the speech signal using minimum number of bits, while preserving the quality of reconstructed speech. For achievement of good compression rates three basic manners in speech coders are used:  representation of frequency-domain characteristic of speech signal  providing waveform coincidence  coder’s optimization according to perceptual properties of the ear According to the coding mechanism there are two types of systems: waveform coders and vocoders. The latter achive better compression rates and work with speech models the most famous type being linear predictive vocoders.