现代电子技术
現代電子技術
현대전자기술
MODERN ELECTRONICS TECHNIQUE
2013年
21期
67-69
,共3页
语音信号处理%端点检测%Mel频率倒谱参数%相关系数%谱熵
語音信號處理%耑點檢測%Mel頻率倒譜參數%相關繫數%譜熵
어음신호처리%단점검측%Mel빈솔도보삼수%상관계수%보적
speech signal processing%endpoint detection%Mel frequency cepstrum coefficient(MFCC)%correlation coeffi-cient%spectrum entropy
为提高低信噪比环境下语音端点检测的准确率，提出了一种基于Mel倒谱参数相似度和谱熵的端点检测算法。首先，提取语音帧的的Mel频率倒谱参数，将前十帧声信号作为背景噪声，然后计算每一帧语音和噪声MFCC的相关系数距离，结合MFCC相似距离与谱熵做综合判决。实验结果表明，在低信噪比环境下此方法相对谱熵法能够提高检测准确率。
為提高低信譟比環境下語音耑點檢測的準確率，提齣瞭一種基于Mel倒譜參數相似度和譜熵的耑點檢測算法。首先，提取語音幀的的Mel頻率倒譜參數，將前十幀聲信號作為揹景譟聲，然後計算每一幀語音和譟聲MFCC的相關繫數距離，結閤MFCC相似距離與譜熵做綜閤判決。實驗結果錶明，在低信譟比環境下此方法相對譜熵法能夠提高檢測準確率。
위제고저신조비배경하어음단점검측적준학솔，제출료일충기우Mel도보삼수상사도화보적적단점검측산법。수선，제취어음정적적Mel빈솔도보삼수，장전십정성신호작위배경조성，연후계산매일정어음화조성MFCC적상관계수거리，결합MFCC상사거리여보적주종합판결。실험결과표명，재저신조비배경하차방법상대보적법능구제고검측준학솔。
In order to improve the accuracy of Voice Activity Detection (VAD) in the low signal-to-noise ratio environ-ment,an algorithm based on Mel frequency cepstrum coefficient(MFCC)similarity and spectrum entropy is proposed. The MF-CC is extracted from each speech frame,and the first ten frames of sound signal are taken as the background noises,then the MFCC correlation coefficient distance of each frame of the voice and noise is calculated,the comprehensive judgment is made in combination with the MFCC similarity distance and spectrum entropy. The experiment result shows that the new algorithm can improve the accuracy of the detection under low SNR environments compared with spectrum entropy method.
