摘要
动态时间规整(Dynamic Time Warping)是语音识别中的一种经典算法,该算法简单有效,在实现孤立词识别系统中得到了广泛的应用。为了提高机器人语音识别系统的识别率和识别速度,文中采用了一种改进的DTW语音识别算法。在MATLAB 7.0环境下,对改进的语音端点检测和改进的DTW算法进行仿真实验,实验证明改进的算法提高了识别率,并且减少了识别所用的时间;将该算法移植到机器人上,在安静的环境下进行试验,结果表明机器人能准确而又快速地识别语音内容。最后,得到了改进的语音识别算法能够有效提高识别率和识别速度的结论。
Dynamic Time Warping （DTW） is a kind of classical algorithm in speech recognition system. Because of the simple and valid, DTW method has been widely used in speech recognition, especially for the recognition of isolated word. In order to achieve the purpose of increasing the recognition rate and recognition speed of robot voice recognition system, an improved DTW speech recognition algorithm is performed here. Improved voice endpoint detection and improved DTW algorithm simulation experiment has been presented in this pro- posed identification algorithm in MATLAB 7.0, the experimental results show that the proposed method can improve the recognition rate and reduce the recognition time. Finally, the algorithm is transplanted to the robot, in a quiet environment the practical application turns out that the robot can quickly identify the content of the speech. In conclusion,the improved speech recognition algorithm can effectively improve the recognition rate and recognition speed.
出处
《计算机技术与发展》
2013年第7期70-72,76,共4页
Computer Technology and Development
基金
辽宁省高校创新团队支持计划项目(LT2010058)
