一种新的双端通话检测方法研究  

Research on a New Detection Algorithm for Double Talk

作　　者：李申[1] 柳玉华[1] 

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《科技广场》2014年第8期93-98,共6页Science Mosaic

摘　　要：为了提升VOIP的通话质量,提高回声抵消系统中双端通话检测的准确性,本文提出一种新的双门限双端通话检测算法。新算法首先利用小波阈值去噪算法对接收的各信号进行去噪处理;然后对去噪处理后的小波子带信号进行动态能量阈值比较,识别出静音帧、活跃帧和发声帧;再对活跃帧与静音帧进行基音周期比较,以判断活跃帧是处在近端发声阶段,还是回声路径的变化。仿真实验证明,新的双门限检测算法表现良好。This paper mainly presents a new algorithm for double-talk detection to improve the quality of VOIP calls and its accuracy in echo cancellation system. This new algorithm firstly de-noises the voice signal via wavelet threshold de-noising algorithm, identifies silent frames, active frames and voice frames by comparing dynamic energy threshold for wavelet sub-band signals after de-nosing. Then the active frames could be determined in a double-talk stage, or changes in the echo path via comparing the pitch determination of active frames and silent frames. Simulation results show that the new algorithm performs well.

关 键 词：双端通话检测 动态能量阈值 基音周期 双门限检测 

分 类 号：TN912.3[电子电信—通信与信息系统]