摘要
针对NAO机器人自身语音识别准确率低问题,提出一种调用云平台服务进行语音识别方法。通过相位补偿的调制域谱减法对声音进行降噪处理,提高语音信噪比,基于双门限判决方法对声音信号进行端点检测,删除噪声帧,保留有话帧,最终生成WAV文件并传输识别,提高了NAO机器人的实用性、功能多样性。实验结果表明,本文算法在低信噪比情况下取得了良好识别效果,具有较强鲁棒性。
Focused on the problems of low recognition accuracy of NAO robot,an approach to speech recognition based on calling cloud platform service is proposed.Through the modulation domain spectrum subtraction method of phase compensation,the noise reduction of the voice is carried out to improve the speech signal-to-noise ratio.The endpoint detection of the sound signal is carried out based on the two-threshold decision method,the noise frames are deleted,the speech frames are retained,and finally WAV files are generated and transmitted for recognition,which improves the practicability and functional diversity of NAO robot.Experimental results show that the proposed algorithm achieves a good recognition effect at low signal-noise ratio and has strong robustness.
作者
陈佳欣
王大东
孙明辰
王晓宇
CHEN Jia-xin;WANG Da-dong;SUN Ming-chen;WANG Xiao-yu(School of Computer Science,Jilin Normal University,Siping Jilin 136000,China)
出处
《佳木斯大学学报(自然科学版)》
CAS
2019年第6期912-916,共5页
Journal of Jiamusi University:Natural Science Edition
基金
吉林省教育厅“十三五”科学技术项目(JJKH20180763K)
关键词
语音识别
调制域谱减法
NAO
降噪
端点检测
speech recognition
modulation domain spectral subtraction
NAO
noise reduction
endpoint detection