摘要
根据压缩感知理论,文中分析了基于压缩感知观测序列语音能量估计的理论基础,并对不同压缩比下语音能量估计的准确度仿真结果做对比,然后将得到的语音能量估计做为语音端点检测的基础,分别对粉红噪声、高斯噪声和汽车噪声情况下不同信噪比的语音进行端点检测,并与基于压缩感知观测序列倒谱距离、传统的奈奎斯特采样中能量的语音端点检测方法做对比,减少了计算量。
Based on the theory of compressing sensing,this paper analyzes the theory of voice energy estimate based on the observation sequence of compressing sensing.The accuracy of energy estimate is evaluated with different compression rate.Then the result of voice energy estimate is used as the basis of speech endpoint detection with pink noise,Gaussian noise and car noise.This paper also makes the comparison with the cepstral distance based on speech observation sequence and the energy endpoint detection based on traditional Nyquist sampling,reducing the amount of computation.
出处
《南京邮电大学学报（自然科学版）》
北大核心
2013年第3期109-113,共5页
Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition
基金
国家自然科学基金(60971129)资助项目
关键词
压缩感知
端点检测
观测序列
语音能量估计
compressing sensing
endpoint detection
observation sequence
voice energy estimate
