摘要
针对一些嵌入式设备在特定的工作场景下无法实现语音识别的问题,提出了一种适用于嵌入式平台语音识别的解决方案。采用百度语音云平台提供的语音识别服务,对自然语言进行语音识别,通过搭建交叉编译环境将录音数据进行压缩编码上传到服务器,交叉编译后下载到嵌入式设备中来完成整个识别过程。通过大量实验数据统计分析,该方案的语音识别准确率可达93%。结果表明:该语音识别方案在正常语速输入下,能够满足工作使用要求,具有较高的准确性和鲁棒性。
Aiming at the problem that some embedded devices can not realize speech recognition under certain working scenes,this paper presents a solution for embedded platform speech recognition. Using Baidu voice cloud platform to provide REST API for natural language speech recognition,cross-compiler environment through the construction of the recording data compression encoding upload to the server,cross-compiler downloaded to the embedded device to complete the entire identification process.Through a large number of experimental data statistical analysis,in the program,the speech recognition accuracy rate is up to 93%. The results show that the proposed speech recognition scheme can meet the requirements of working under the normal speech speed,and has high accuracy and robustness.
出处
《信息技术》
2017年第11期117-120,共4页
Information Technology
关键词
语音识别
嵌入式系统
自然语言
百度语音
speech recognition
embedded system
natural language
baidu voice
