A Nonorthogonal Joint Block Diagonalization Algorithm Avoiding Degenerate Solutions

摘要

The joint block diagonalization algorithm is instrum ental to the convolution blind signal separation. However, the nonorthogonal joint block diagonalization al-gorithm based on a Weighted least-squares (WLS) crite-rion sometimes converges to the degenerate or trivial so-lution. It is simple to avoid the trivial solution by adding some constraint on the de-mixing matrix. But the degen-erate solution is hard to be excluded. This paper probes the phenomenon of the degenerate solution. By adding a new penalty function to the WLS criterion, an efficient algorithm which can avoid the degenerate solution is de-veloped. Through the theoretical derivation and numerical simulation, we can make the conclusion that the proposed algorithm is more robust and not sensitive to the square or non-square m ixing m atrix.