基于定位标签的时间规整方法

Time-Aligned Method based on Positioning Tag

为了评估话音通信系统的性能,通常需要把输出语音信号进行时间规整,然后才能比较输入语音信号与输出语音信号的相似度。传统的语音信号时间规整算法适应性差,当环境噪声增大时,时间规整误差会导致系统性能急剧下降。因此,提出了一种基于定位标签的时间规整算法,大大提高了低信噪比语音信号的时间规整精度。实验仿真及工程实践表明,当信噪比为-20 d B时,提出的算法仍能达到非常高的估计精度,远远超过传统方法的时延估计精度,具有很强的抗干扰能力。

In order to evaluate the performance of a voice communication system, it is usually necessary to time warp the output speech signal before comparing the similarity of between the input speech signal and the output speech signal. The traditional speech signal time warping algorithm has poor adaptability, and when the ambient noise increases, the time-warping error can cause a sharp drop of system performance. A time-regulating algorithm based on positioning tags is proposed, thus to greatly improve the time warping precision of low SNR speech signals. Experimental simulation and engineering practice indicate that when the signal-to-noise ratio is -20 dB, the proposed algorithm could still achieve very high estimation accuracy, far exceeding the delay estimation accuracy of traditional method, and has strong anti-interference ability.