指数形式的二次相关GCC-PHAT-β时延估计算法  

Exponential Quadratic Correlation GCC-PHAT-βDelay Estimation Algorithm

作　　者：李皖婷 李登峰[1] 张修靖 张天婷 黄杉杉 LI Wan-ting;LI Deng-feng;ZHANG Xiu-jing;ZHANG Tian-ting;HUANG Shan-shan(School of Electronics and Control Engineering,Chang􀆳an University,Xi􀆳an 710064,China)

机构地区：[1]长安大学电子与控制工程学院,西安710064

出　　处：《科学技术与工程》2022年第33期14779-14785,共7页Science Technology and Engineering

基　　金：中国博士后科学基金(2016M600756);陕西省重点产业创新链基金(2018ZDCXL-GY-05-03-02)。

摘　　要：针对现有的时延估计算法在低信噪比下精度低、时延估计误差大的问题,提出一种指数形式的广义互相关-相位变换(generalized quadratic cross-correlation phrase transformation-β,GCC-PHAT-β)算法。通过构造加有高斯白噪声的两路单频正弦波信号,分别在不同信噪比下对GCC-PHAT-β算法、二次相关GCC-PHAT-β算法以及改进算法进行实验仿真,验证三种算法在不同信噪比下的峰值凸出程度;进一步选用男声音频作为信号源,比较3种算法在进行时延估计时所产生的误差大小。仿真结果表明:信噪比为-10~10 dB时,改进算法的相关函数的最大峰值明显凸出于GCC-PHAT-β算法和二次相关GCC-PHAT-β算法,其时延估计误差相较于其他两种算法也显著减小。可见指数形式的二次相关GCC-PHAT-β算法在信噪比为-10~10 dB时具有一定的优势。Aiming at the problems that the existing delay estimation algorithms have low accuracy and large delay estimation errors at low signal-to-noise ratios,an exponential quadratic correlation generalized cross correlation-phrase transform-β(GCC-PHAT-β)algorithm was proposed.By constructing two single-frequency sine wave signals with Gaussian white noise,the GCC-PHAT-βalgorithm,the quadratic correlation GCC-PHAT-βalgorithm and the improved algorithm were simulated under different signal to noise ratio(SNR),respectively.The algorithm and the improved algorithm were simulated to verify the peak bulge degree of the three algorithms under different SNR.The male voice audio was further selected as the signal source,and the errors generated by the three algorithms in the time delay estimation were compared.The real results show that when the SNR is-10~10 dB,the maximum peak value of the correlation function of the improved algorithm is obviously higher than that of the GCC-PHAT-βalgorithm and the quadratic correlation GCC-PHAT-βalgorithm.It is also significantly reduced for the other two algorithms.It can be seen that the quadratic correlation GCC-PHAT-βalgorithm in exponential form has certain advantages when the SNR is-10~10 dB.

关 键 词：时延估计 低信噪比 改进广义二次相关算法 凸出峰值 指数运算 

分 类 号：TN912[电子电信—通信与信息系统]