An optimized fast codeword search algorithm for vector quantization

Vector quantization is a popular data compression method with a widespread use in signal processing and pattern recognition applications. However its computational demand is increased tremendously as the size of date set and the dimension of samples grow larger. Here we have proposed a new fast-search algorithm to enhance the computational cost and the execution time of the conventional VQ method. This algorithm relies on the Chebyshev distance for distortion measurement and Triangular Inequality Elimination (TIE) for codeword elimination. The proposed algorithm was employed for the recognition of isolated spoken Persian digits. Results show that this algorithm can remarkably reduce the number of calculations while its encoding error is closed to the full search algorithm.