DEVELOPMENT OF SPEECH COMPRESSION ALGORITHMS

Abstract

Speech compression is the technology of converting human speech into an efficiently encoded representation that can later be decoded to produce a close approximation of the original signal. Basically compression of the speech signal is to save the bandwidth and reduce the storage capacity of the signal. This dissertation report presents study of different speech compression algorithms including waveform coding, parametric coding and the hybrid coding to compress speech signals. Waveform coding includes the Sub band coding algorithm while parametric coding which is a model based coding covers the Linear Predictive Coding algorithm. In Sub Band Coding algorithm the given frequency band of the input speech signal is get divided into the frequency sub bands by the low pass and high pass filters. The output of each filter is then downsampled and encoded. The signal in each band is encoded using a different number of bits, depending upon the statistics of the signal. In speech coding applications, the higher frequency bands will require fewer bits than the lower frequency bands. The filters in the coder and decoder must be designed carefully so that perfect reconstruction is achieved. The Linear Predictive vocoder is a model based coder. The parameters of the model are quantized and transmitted to the receiver It is a digital method for encoding an analog signal in which a particular value is predicted by a linear function of the past values of the signal. Under normal circumstances speech is sampled at 8000 samples/second with 8 bits used to represent each sample. This provides a rate of 64000bits/second or we can say that the raw data or input speech signal have bit rate of 64 kbps. In this report this raw data get implemented and compared by the use of both SBC and LPC techniques, which shows the approximately compressed results upto 16 kbps and 2.4 kbps respectively .