论文部分内容阅读
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。
Discusses a typical telephone speech recognition system design and implementation. First, the overall model of the system is proposed, followed by the overall structure of the system and the main modules of the system. The speed optimization of acoustic layer recognition and telephone channel compensation are introduced in detail. The fault tolerant algorithm of language model and the fault-tolerant understanding model based on semantic dependency are also introduced. The module performance of the system is experimentally tested.