电话语音识别系统

Telephony Speech Recognition System

讨论了一个典型的电话语音识别系统的设计与实现。首先提出了系统的整体模型，之后给出了系统的整体结构和系统的主要模块。较详细地介绍了声学层识别的速度优化和电话信道补偿问题，以及语言模型的容错算法和基于语义依存关系的容错理解模型，并对系统各层次模块性能进行了实验测试。

This paper discusses the implementation of a typical Telephony Speech Recognition System.First the framework of the integrated model is presented and then the structure of the system and the main modules' are given.Some enhance technologies such as the improvement of the acoustic recognition speed,the compensation of the telephone channel effect,the tolerance of language model and the understanding model based on semantic dependency relations are discussed in detail.