摘要： 介绍了一种基于联合得分的实时孤立词语音识别系统。在识别阶段，通过内插向量得到两种改进的DTW 算法，并且和基本的DTW联合起来对语音进行识别，得出各自的识别结果，最后把所有的结果联合起来得到最终的识别结果。通过在TI46 语音库和实时运行的实验表明，系统在低信噪比的环境下和实时运行的环境中，都可以获得比一般只应用DTW 算法的识别系统更高的识别率。

关键词: 语音识别；端点检测；联合得分

Abstract: A real-time speech recognition system is presented based on a combination score. In the process of recognition, two improved DTWs are attained by vector-interpolation. The two improved DTW recognizers are then combined with original DTW recognizers to recognize speech.Based on the experiments with TI46 database and in real-time test, this system can achieve higher recognition rate than solely DTW-based recognition system in low SNR environments and in real-time environment

Key words: Speech recognition; Endpoint detection; Combination score