论文部分内容阅读
基于到达时差(TDOA)的声源定位是声音信号处理中一个非常活跃的研究方向,具有广泛的应用。由于实际环境中的噪声和混响的影响,声源定位是一个具有挑战性的课题。针对具有噪声和混响的室内环境提出了一种鲁棒声源定位方法。首先,通过麦克风阵列活跃性的统计建模确定包含声源信息的有效帧。接下来,基于随机采样一致性算法提出了一种鲁棒算法使用TDOA对声源进行定位。有效帧检测算法是一种无监督的方法,有效地利用了多通道信息,能够方便地应用于不同环境中的音频数据;同时,基于随机采样一致性的算法具有很强的鲁棒性,能够在具有噪声和混响的不利环境中对声源进行准确定位。实验表明所提出的方法能够显著地减少声源定位的偏差而且具有较小的标准差。
Sound source localization based on arrival time difference (TDOA) is a very active research direction in sound signal processing and has a wide range of applications. Due to the noise and reverberation in the actual environment, sound source localization is a challenging issue. Aiming at indoor environment with noise and reverberation, a robust sound source localization method is proposed. First, a valid frame containing sound source information is determined by statistical modeling of microphone array activity. Next, a robust algorithm based on stochastic sampling consistency algorithm is proposed to locate the sound source using TDOA. Effective frame detection algorithm is an unsupervised method, effectively using multi-channel information, can be easily applied to different environments in the audio data; at the same time, based on random sampling consistency algorithm has strong robustness, able to Accurately locate the sound source in adverse environments with noise and reverberation. Experiments show that the proposed method can significantly reduce the source localization error and have a smaller standard deviation.