人机交互系统多维语音信息识别方法

Multi-Dimensional Speech Information Recognition Method of Human-Computer Interaction System

利用现有方法对多维语音信息进行识别时,没有对多维语音信息进行相关预处理,存在语音识别率低、识别效率低和识别准确率低的问题。提出人机交互系统多维语音信息识别方法,通过预加重、分帧加窗和端点检测对多维语音信息进行预处理,消除由于人类发声器官本身和环境引起噪声等因素对多维语音信号质量产生的影响,平滑多维语音信息,提取多维语音信息的特征参数,通过关联规则重组方法对人机交互系统多维语音信息的特征参数进行融合,获得多维语音信息特征,完成人机交互系统多维语音信息的识别。实验结果表明,所提人机交互系统多维语音信息识别方法的语音识别率较高、识别效率较高、识别准确率较高。

Due to the lack of multi-dimensional speech information preprocessing, the traditional method ofmulti-dimensional speech information recognition has many defects, such as low speech recognition rate, low recog-nition efficiency and low recognition accuracy. In this regard, we designed a multi-dimensional speech informationrecognition method for human-computer interaction system. Multi-dimensional speech information was preprocessedby pre-emphasis, frame windowing and endpoint detection, in order to eliminate the influence of noise caused by hu-man vocal organs and environment on the quality of multi-dimensional speech signal. Multi-dimensional speech in-formation was smoothed to extract feature parameters of multi-dimensional speech information. Association rules reor-ganization method was applied to fuse the feature parameters of multi-dimensional voice information of human-com-puter interaction system, thus obtaining the features of multi-dimensional voice information. Eventually, the recogni-tion of multi-dimensional voice information of human-computer interaction system was completed. The experimentalresults show that the method has high recognition rate, high recognition efficiency and high accuracy.