一个应用线性判别分析的关键词识别方法

A Keyword Spotting Method Applying LDA

关键词识别是一种特殊的语音识别技术,它旨在连续语音中通过对特征矢量的判别检测出感兴趣的词。论文给出一种新的检测方法,应用线性判别分析(LDA)对语音特征参数进行降维,使得分类更加清晰。通过实验表明了该方法提高了系统的性能。

Keyword spotting is a special technique in speech recognition task.Its objective is finding the interesting words by detecting feature vectors.In this work,a new method using a Linear Discriminant Analysis(LDA)which reduces the dimension of feature vectors is presented.It makes classification clearer.Experimental results indicate that this method has improved the performance of system.