Fast codeword search technique for the encoding of variable-rate vector quantisers

The authors present a fast search algorithm for the encoding of variable-rate vector quantisers (VQs). The optimal encoder for the variable-rate full-search VQs considers both the squared distance and channel index length for the search of codewords. Therefore, the existing fast search algorithms for the fixed-rate full-search VQs might not be able to effectively reduce the computational complexities for the variable-rate VQs. An algorithm is presented which solves this problem using a two-stage search process. A large number of undesired codewords are removed at the first stage of the search with a small computational overhead. At the second stage, the optimal codeword are searched among the remaining codewords using the partial distance search technique in the transform domain. Simulation results show that the algorithm is more effective than other fast search algorithms for the variable-rate VQs.