论文部分内容阅读
近年来,基于时延估计(TDOA)的声源定位系统得到了广泛的关注和研究。传统的相关运算时延估计法在噪音较强情况下,得到的时间差误差较大。本文在讨论短时能量法、过零率法、传统双门限等端点检测方法的基础上,研究通过获得声源到达各传感器的精确起始时间,从而得到时间差,提出一种改进的谱熵—双门限端点检测算法,并在此基础上加以改进。实验表明,该端点检测法在能快速准确判别信号起止点,并具有简单、准确的特点,比较适用于DSP等小型系统中。
In recent years, time-delay estimation (TDOA) based sound source localization system has received extensive attention and research. The traditional correlation delay estimation method in the case of strong noise, the error of the time difference obtained. Based on the discussion of the short-time energy method, the zero-crossing rate method and the traditional two-threshold method, this paper studies the accurate start time of the sound source reaching each sensor to obtain the time difference, and proposes an improved spectral entropy- Double threshold endpoint detection algorithm, and on this basis to be improved. Experiments show that the endpoint detection method can quickly and accurately identify the start and end points of the signal, and has the characteristics of simplicity and accuracy, and is suitable for small systems such as DSP.