基于压缩感知的麦克风阵列声源定位算法

Compressed sensing-based sound source localization algorithm for microphone array

为了提高麦克风阵列在高混响、低信噪比环境中的定位性能,提出了一种基于压缩感知的声源定位算法.该算法将声源定位问题转化为稀疏信号的重构问题,将不同位置的房间冲激响应作为特征以构建字典.首先,将麦克风接收信号转换至频域,从具有较高能量的频点中求得一组扩展的频域声源信号矢量,该矢量中包含了声源的位置信息.然后,在频域中整合这些扩展的声源信号矢量,使声源的位置信息更突出,矢量中最大元素所对应的空间位置即为声源的位置估计.仿真实验结果表明,与相位变换加权的可控响应功率(SRP-PHAT)定位算法相比,所提算法的定位成功率更高,对混响的鲁棒性更强,更适合高混响低信噪比环境中的声源位置估计.

To improve the sound source localization performance of microphone arrays under the conditions with high reverberation and low signal-to-noise ratio（ SNR）,a compressed sensing-based sound source localization algorithm is proposed. In the proposed algorithm,the problem of sound source localization is converted to the reconstruction problem of sparse signal,and the room impulse responses at different locations are treated as the features used to construct the dictionary. First,the received signals of the microphone array are transformed to the frequency domain,and a set of extended source signal vectors in the frequency domain,which contain the location information of the sound source,are calculated from the frequency components with higher power. Then,the extended source signal vectors are integrated in the frequency domain to enhance the location information of the sound source,and the spatial location corresponding to the maximum element of the integrated vector is the location estimation of the sound source. The simulation results show that compared with the steered response power-phase transform（ SRP-PHAT） localization algorithm,the proposed algorithm has a higher localization rate,and is more robust against reverberation and more suitable for location estimation under the conditions with high reverberation and low SNR.