基于相位差复指数变换的传声器多声源定位

Multiple sound source localization of microphones based on complex exponential transform of phase differences

为了提高多声源定位的性能并且避免针对高频段相位差的去卷绕处理,提出了一种基于相位差复指数变换的传声器多声源定位算法.首先,挑选出信噪比较大的频点,以提高算法对噪声的鲁棒性,并对被挑选出的频点作相位差复指数变换;然后,基于语音信号在时-频域的稀疏性,将被挑选出的频点聚类至各声源;最后,利用各声源包含的频点构建代价函数,使得代价函数最小的假设时延差即为估计的声源时延差.仿真结果表明,该算法充分利用了高频段的相位信息,无需对高频段相位差进行去卷绕处理.相比广义硬聚类算法,该算法收敛速度更快,定位成功率更高,均方根误差更小.

To improve the performance of multiple sound source localization and avoid the phase-unwrapping process at high frequencies,a localization algorithm for multiple sound sources of two microphones based on complex exponential transform of phase differences is proposed.First,the reliable frequency points with high signal-to-noise are selected to enhance the robustness of the proposed method against noise.The inter-channel phase differences（IPDs） of the selected frequency points are converted with complex exponential transform.Then,the selected frequency points are clustered into groups corresponding to multiple sources based on speech＇s sparse attribute in time-frequency domain.Finally,the assumed time-delay difference which minimizes the cost function building on the frequency points of each source is the estimated time-delay difference of the source.The simulation results show that the proposed algorithm takes full advantage of the high-frequency phase information without phase-unwrapping process for IPDs at high frequencies.Compared with the generalized hard clustering algorithm（GHCA）,the proposed algorithm has a faster convergence speed,and obtains a higher percentage of successful estimates and a lower root mean square error.