录音时间识别的并行匹配算法研究  

Algorithm Study of Parallel Matching in Recording Time Identification

作　　者：吉建梅 王学强[1] 包永强[1] 

机构地区：[1]南京工程学院通信工程学院,南京211167

出　　处：《信息化研究》2014年第4期36-40,共5页INFORMATIZATION RESEARCH

基　　金：国家自然科学基金(51075068);江苏省大学生实践创新训练计划项目(201311276016Z)

摘　　要：录音时间识别是语音取证的关键技术和方法。传统的匹配方法速度较慢,准确率不高,文章提出了一种新的录音时间识别的匹配方法,首先将提取到的待匹配的音频中的电网频率信号分割成三段与电力线中电网频率各帧进行并行匹配,得到两者的相关程度,并与阈值进行比较,判断出相关性最大的帧数,进而得出音频信号的录音时间,从而作为法庭和其他行政部门认可的证据。实验结果表明,该并行匹配算法能够完成高性能的并行计算,与传统的方法相比,该方法采用并行机制,提高了匹配效率,识别率达到了92.5%,速度更快。Recording time voice recognition is evidence of key technologies and methods.Taking into account the slower traditional matching method,the accuracy rate is not high,we propose a new method for identifying the recording time of the match,the first match will be extracted to the grid frequency audio signal is divided into three sections and power lines parallel grid frequency to match each frame to obtain the degree of correlation between the two,and with the threshold value to determine the correlation between the maximum number of frames,and then come to the audio signal recording time,so as recognized by the courts and other administrative departments of the evidence.Experimental results show that the parallel matching algorithm to complete performance parallel computing,compared with the conventional method using aparallel mechanism,improve the efficiency of the matching accuracy rate reached 92.5%,which is faster than others.

关 键 词：时间识别 语音取证 电网频率 并行匹配 相关程度 

分 类 号：TN912.34[电子电信—通信与信息系统]