基于“增加树”的语音识别搜索网格的优化设计

Speech Recognizing Network Optimization Design Based on "Growing Tree"

基于传统的语音识别搜索网格的分析,提出了一种基于“增加树”法和上文相关声学模型方法的识别搜索网格优化,该方法对于人名查询搜索网格中姓氏的大量冗余能够起到较好的优化作用,该识别搜索网格优化设计主要针对一个人名查询应用的识别内核开发提出来的。

After analysis of traditional method,a kind of recognizing network optimization based on ＂growing tree＂ method and left-context-dependent model is presented to reduce the redundant nodes and transitions in network,it is designed for the core of a name retrieval system.