摘要： 为提高语音识别系统的识别速度而采用两级识别,是一个行之有效的方法。采用连续语音的音素结构(PS)作为第一识别,是其中重要方法之一。本文介绍了根据普通话具有“四声”的特点,利用语音中的声母、韵母、以及韵母部分的“四声”来形成描述该语音的结构参数(PS)的方法。文章分析了利用音素划分,以及利用基频参数进行“四声”识别等问题。实践结果表明,利用这种算法进行普通话四声识别是简易可行的,并对提高识别系统的速度有明显的效果。

Abstract: In order to increase the recognition speed a two-stage recognition system is used. As the first stage, the phoneme structure (PS) for the Chinese word or phrase based on the "four tone" of Chinese pronunciation is available. The issues on the segmentation of the phoneme and the decision of "four-tone" according to the short time basic frequency are discussed in this paper. Facts indicate that this strategy is efficient for the "four tone" decision algorism and theretore for having higher speed and accuracy for a speech recognition system.