基于EDFE的顽健语音识别方法中测试环境特征的自适应

Adaptation of Testing Environment in Robust Speech Recognition Based on EDFE

作者曾提出一种基于环境特征判别学习的顽健语音识别方法EDFE，该方法基于最小分类错误准则迭代地学习环境特征。从训练数据中学得的环境特征不可能完全反映测试环境的特点，因此研究测试环境的自适应方法，有利于提高语音识别系统性能。文章讨论EDFE中一种测试环境的自适应方法。

A robust speech recognition method,based on discriminative learning of environmental fetus,EDFE wasproposed in our previous work,in which the environmental features are iteratively learned according to Minimum Classilication Error (MCE) criterion. Generally,the environmental features learned from the training samples are not suitableto the testing environment. Thus,using adaptation of testing environment is helpful to improving the performance ofspeech recognition. In this paper,an adaptation method of testing envimnment in EDFE is discussed.