摘要
在SEED-DEC5502DSP嵌入式系统开发平台上实现了一个面向非特定人的孤立词语音识别系统,和传统的基于特定人的语音识别系统相比,该系统无需用户训练,易于使用。系统采用改进的基于语音对数域能量变化率的实时端点检测算法,只对检测的有声段语音进行特征提取,从而减少了要处理的语音帧数;提出了改进的共享声学单元状态发射概率共享的解码策略,进一步降低了计算负担。实验表明系统在100词条的情况下识别率达到98.1%,识别时间为1.03倍实时。
An embedded speaker-independent isolated word speech recognition system is designed and realized in the SEED- DEC5502 EVM platform.Compared with the speaker-dependent system,the speaker-independent recognition technique cannot requires training by the users and easy to use.With the help of a modified real time Voice Activity Detection algorithm（VAD） based on the log-energy acceleration associated with voice onset,we only perform feature extraction to the active voice and decrease the frames of processing.To further decrease the computational loads,A modified decoding strategy based on the share the uint states emission probabilities （SUSEP） is also presented.Test on 100 words vocabulary shows that system provides a recognition accuracy rate of 98.1% using only 1.03 times of real time.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第13期194-196,共3页
Computer Engineering and Applications
基金
河北省教育厅自然科学指令计划(No.2005340)
河北省科技厅科技发展规划指导计划(No.052135147)。
关键词
语音识别
嵌入式系统
端点检测
发射概率
speech recognition
embedded system
speech endpoint detect
state emission probabilities
