基于波束形成器输出比的自适应语音分离方法

Automatic Adaptive Speech Seperation Method Based on Beamformer- output- ratio

运用波束形成的方法能够有效地从干扰和噪声中提取所需的语音信号,如最小方差无失真响应的方法。然而,为了避免应用波束形成器进行语音分离时所产生的信号抵消现象,当目标讲话者处于活动状态的时候,该波束形成器的适应性应处于暂停状态,这就需要对目标讲话者的语音活动状态进行分类。基于这种思想,首先从信号模型和系统模型出发,介绍了"波束形成器输出比"这个新量的相关概念和基于波束形成输出比的"BOR-VAC"新方法。而后对窗口值、阈值的选取和VAC模块的实现过程进行了详述。最后通过仿真实验实现并验证了这种基于波束形成器输出比的自适应语音分离新方法的高效性和可靠性。

The beamforming method can be used to effectively extract the desired speech signal from interference and noise. However,to avoid signal cancellation,the classification for the speakers' voice activity status is required. In this paper,we study and implement a new method based on beamformer-output- ratio,and construct an automatic adaptive beamforming system to implement speech separation.The simulation verifies the reliablity of the algorithm.