A new efficient method for permutation and scaling ambiguity of blind source separation signal blocks

This paper deals with the permutation and scaling ambiguities problem of blind source separation (BSS) in the case where continuously mixing signals are split in time and processed block by block. When tying the separated components at each signal blocks, the original source signals can't be recovered correctly because of the inherent permutation, amplitude and phase ambiguities of BSS. Inspired by the permutation method for separated source signal blocks in time domain, a new approach is proposed to eliminate the ambiguities by artificially setting contrast blocks for each adjacent mixing signal blocks. The separated signal waveform difference in time domain between contrast and corresponding adjacent blocks is utilized to adjust the permutation and scaling parameters. This new method is similar in spirit to the separation and reconstruction method for the mixing and long duration communication signals, but differs in the fact that it is more efficient in terms of computational speed than the latter, which is significantly striking with large number of sources and signal blocks. Furthermore, when the block size and corresponding length of overlapping signals are chosen appropriately, this new method is more efficient than the permutation method in terms of separation quality and computational speed. Realistic experiments based on the wireless communication system validate the performance of this new method, together with corresponding comparison and analysis performed.