麦克风阵列声源定位可为复杂环境下的说话人空间位置估计问题提供一种有效的解决方案。该文基于粒子滤波框架,提出了一种加权子空间拟合声源定位与跟踪方法。该方法将窄带子空间拟合算法的代价函数推广至宽带情形,构建了一种适用于宽带语音信号的似然函数,并结合说话人的运动模型估计声源的位置。计算机仿真与实测结果验证了该方法的有效性。
Sound source localization using microphone array provides an effective solution to speaker tracking problem under adverse environments. A new method based on weighted suhspace fitting is presented for sound source localization and tracking within the framework of particle filtering. The cost function of weighted subspace fitting algorithm is extended to the wideband case, and the location of speaker is estimated using dynamical model of speaker motion and the likelihood function model based on the wideband cost function. The results of both simulations with synthetic data and experiments with real-world data show that the proposed method has good performance.