High Computational Performance in Code Exited Linear Prediction Speech Model Using Faster Codebook Search Techniques

The most commonly used coder for producing good quality speech at rates below 10 kbits/s is code excited linear prediction (CELP). In this time domain analysis-by-synthesis coder, the excitation signal is chosen by attempting to match the reconstructed speech waveform as closely as possible to the original speech waveform by searching through a large vector quantizer codebook. The closest matching excitation signal (taken from code book) along with pitch, linear prediction coefficients (LPC) and pitch lag are typically transmitted from the encoder side in order to reconstruct the input speech signal. Codebook search is identified as the computationally most complex module and several methods to implement the same are described along with complexity reduction due to each method