For signal enhancement, beamforming remains to be an essential technique for many applications. In the design process, the microphone locations are prescribed and the signal from a target location is being enhanced. While the filter coefficients can be readily optimized, it is found that the signal enhancement capability depends significantly on the array configuration. Therefore, it is advantageous to consider both filters and microphone positions as design variables. In this paper, this problem is addressed. We formulate the beamformer design problem as a non-linear least square problem and propose Gauss-Newton algorithm to update both filters and configuration simultaneously during iterations. We illustrate by several designs to demonstrate the effectiveness of the proposed method.
展开▼