Three-particle cooperative optimization method applied to vector quantization-based speaker recognition

Abstract

The invention relates to a three-particle cooperative optimization method applied to speaker recognition, which is an optimization design method for a vector quantization speaker codebook model and comprises the following steps: dividing an initial population of six particles into two sub-populations, wherein each particle represents a codebook and each sub-population consists of three particles named three-particle; adopting different particle update parameters by two three-particles so as to realize global exploration and local exploration; executing updating of speed and position of PSO and operation of an LBG algorithm with an iteration number of three by the particles in each iteration; when the iterations reach mixing update times, mixing and dividing the particles into new three-particles to realize global information exchange and cooperative evolution; when a condition of the maximum iteration number of the initial population is met, selecting two particles from the two three-particles to continue searching until the maximum iteration number of the elite particles is achieved; and using the optimized one as the speaker codebook model. The three-particle cooperative optimization method applied to the speaker recognition better solves the problem that an initial codebook influences the optimization result, and obviously improves the performance of short-voice speaker recognition.

Description

technical field [0001] The invention relates to the technical field of speech recognition, and more specifically, relates to a three-particle cooperative optimization method applied to speaker recognition based on vector quantization. Background technique [0002] Speaker recognition is a biometric technology that is urgently needed for the application of information technology in existing communication networks. Judicial field (such as monitoring and identification of criminals), security field (such as airport access control system control), information service field (such as automatic information retrieval or e-commerce), etc. [0003] The speaker model is the core of high-performance speaker recognition. Since the late 1980s, speaker recognition has entered a new and thriving development due to the establishment of speaker models based on vector quantization, probability statistics models, and artificial neural networks. period. As a typical optimization problem, model...

AI Technical Summary

Problems solved by technology

If the initial codebook is not well selected, it will affect the codebook performance and program operation time, thus affecting the speaker recognition performance

[0049] Although the codebook design quality of the particle pair collaborative optimization algorithm has been greatly improved compared with the traditional LBG algorithm, there is still a certain sensitivity to the selection of the initial codebook, which may fall into the local optimum, and cannot guarantee that the global optimum can be found. Ucodebook

Images