多噪声环境下声音的身份鉴定技术研究与仿真  

More Under the Environment of the Identity of the Noise Voice Identification Technology Research and Simulation

作　　者：马仲海[1] 

机构地区：[1]西北民族大学数学与计算机科学学院,甘肃兰州730030

出　　处：《计算机仿真》2012年第5期396-399,共4页Computer Simulation

摘　　要：研究多噪声背景下基于声音的身份准确识别问题。当个体声音混淆在其它声音背景下的时候,个体独特的语音特征由于混有噪声特征,导致自身特征发生退化。造成的传统语音特征进行识别的方法,由于无法克服"噪声"干扰下语音特征的退化,造成识别准确性不高。提出了根据语音特征概率决策的混合语音识别方法。采用对混合语音进行独立的有效特征检测,根据特征概率决策算法将语音特征进行归属概率识别,并通过特征归属度计算完成身份识别,克服由于语音特征混合,造成语音特征退化造成的识别不准问题。实验结果表明,提出的方法能够对混合的语音特征进行有效的识别,提高了识别的准确度。Study background noise based on sound identity accurately identify problems.When the individual voice confusion in the other voice background,individual unique speech characteristics due to the mixed with the noise characteristics,leading to their own characteristics occur degradation.By the characteristics of the traditional based on the speech recognition methods and,unable to overcome ＂noise＂ speech characteristics of the degradation of disturbance,cause recognition accuracy is not high.Was put forward based on the phonetic features of the mixed speech recognition probability decision-making method.Through the mix of voice independent effective feature detection,according to the characteristics of decision-making algorithm will phonetic features probability belonging to identify probability,through the calculation of ownership features complete identity recognition,and overcome speech characteristics due to the mix,causing the phonetic features of the result of degradation to identify problems.The experimental results show that this method can mix the phonetic features to effectively identify and improve the accuracy of the recognition.

关 键 词：语音识别 混合语音 特征归属决策 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]