Overcoming block permutation problem in frequency domain blind source separation when using AuxIVA algorithm

Auxiliary function based independent vector analysis (AuxIVA) is a recently proposed method for blind source separation which can guarantee the monotonic decrease of the objective function at each update. However, the traditional source dependency model used by AuxIVA may cause a block permutation problem. In this reported work, the reason for this problem is confirmed, and analytically experimental results verify that this problem happens regularly. Therefore, an improved source dependency model is adopted in AuxIVA. Experimental results show that this approach can overcome the block permutation problem and converge faster than AuxIVA on speech sources mixed in a reverberant room environment.