Analysis and improvement of the partial distance search algorithm
Acoustics, Speech, and Signal Processing, 1993. ICASSP-93., 1993 IEEE International Conference (1993)
摘要
This paper analyzes the Partial Distance Search algorithm (PDS) introduced by Bei and Gray for reducing the computational complexity of the nearest neighbor search in vector quantization. In particular a detailed analysis is performed of the computational savings that can be obtained by minor modifications to this algorithm. Since the efficiency of the search is related to the elimination of a reference vector if its partial accumulated distance from the first n components of the input vector is larger than the current minimum distance, a Dynamic Programming procedure is proposed that automatically determines how often the comparison with the current minimum distance has to be done in order to minimize the expected global cost of the search. The number and position of the comparisons within the distance evaluation loop depends on the ratio of the cost of a comparison operation with respect to that of the partial distance evaluation. It is shown that the two costs are comparable for RISC processors, and a 25% speedup with respect to the PDS algorithm is reported for 24 dimension feature vectors used in a Continuous Density HMM system with 16 Gaussian mixtures per state.
更多查看译文
关键词
computational complexity,dynamic programming,hidden Markov models,reduced instruction set computing,search problems,speech recognition,vector quantisation,Gaussian mixtures,RISC,computational savings,distance evaluation loop,dynamic programming,feature vectors,global cost,hidden Markov model,nearest neighbor search,partial distance search algorithm,vector quantization
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络