基于GCC-PHAT的改进近场声源定位算法

Improved acoustic source localization algorithm based on GCC-PHAT in near-field situation

结合GCC-PHAT与SRP搜索的算法被广泛应用于声源定位系统,但其定位结果随信噪比降低而恶化严重,且难以估计窄带声源。针对这些问题,提出一种用于近场模型的改进算法。首先计算阵列信号求和功率谱的峰均比,再经对数变换和设置噪声抑制阈值等步骤,得到反映频点信噪比的加权因子,将其应用在GCC-PHAT中,并进一步优化SRP搜索策略。大量仿真结果表明:改进的算法具有强的抗噪能力,且对宽带和窄带声源信号均有良好的估计效果,很适合实时处理。

GCC-PHAT combined with SRP searching algorithm had proved to be ineffective on narrowband sound source and the localization results deteriorated severely with the decreasing of SNR. To solve these problems,proposed an improved algorithm for near field model which introduced a weighting factor by calculating the PAPR of received signals in the frequency domain, and then make logarithmic transformation to compress the dynamic range,as well as the establishment of a threshold to suppress noise,etc. ,and further optimized the sound source position searching strategy. A lot of simulation results showed that the improved algorithm has strong ability to resist noise and have good estimation effects on both broadband and narrowband sound sources,it＇s very suitable for real-time processing.