Essay On Speech Compression

Speech Compression Using Wavelet Packets Tree Nodes LPC Encoding and Best Tree Encoding (BTE) Features

ABSTRACT
Over the past several years there has been considerable attention focused on compression and enhancement of speech signals. This interest is progressed towards the development of new techniques capable of producing good quality speech at the output. Speech compression is a process of converting human speech into efficient encoded representations that can be decoded to produce a close approximation of the original signal. This paper deals with the problem of speech compression. It proposes the application of Best Tree Encoding (BTE)[1] to encode the wavelet packet Best Tree Structure into a vector of 4 elements. Tree node data coefficients

LPC technique is a developed algorithm used in speech analysis for many years. Its basic idea comes from a model representing the resonances of the human vocal tract. In general, speech sounds are produced by acoustic excitation of the vocal tract. During the production of voiced sounds, the vocal tract is excited by a series of nearly periodic pulses generated by the vocal cords. With unvoiced sounds, the excitation is provided by air passing turbulently through constrictions in the tract [14-20]. Speech signal is non stationary signal. To deal with non-stationary signal the segmentation technique is used to convert the long duration of speech into small segments for which we can consider it stationary signal. This process is called framing the signal.

After the speech signal is segmented, the p autocorrelation coefficients are estimated, where p is the linear predictor order. The autocorrelation function can be estimated using the biased or unbiased autocorrelation algorithms [21-24]. Once autocorrelation coefficients p are evaluated from each segment, the signal at time n can be rewritten as a linear combination from the past samples of the input signal: =-(a_1 s(n-1)+a_2 s(n-2+⋯+a_p (n-p))