Research on multiple-classifiers fusion for speech recognition
基于多分类器融合的语音识别方法研究

Fusion of multiple classifiers can be integrated superiority of the each classifier, and better recognition effect can be achieved. At present, the multi-classifier fusion is a hot topic in pattern recognition. In a speech recognition system, design of the classifier is the key point to the superiority. In this paper, a novel speech recognition approach based on multi-classifier fusion is proposed. SVM, RBF network and bayes net are fused by weighted voting strategy, and the weight is calculated according to the validation sets from sample database. The experiment results show that the performance of multiple-classifiers fusion is better than single classifier, and the method proposed in this paper is effective.