基于非正交对角化算法的非平稳信号盲分离

运用分段平稳的思想处理非平稳信号,对每一子分段信号进行计算,得到相关矩阵群,并运用非正交对角化算法(alternatmg columns diagonal centers,ACDC)求得分离矩阵的逆阵,从而在混合信号中分离出源信号.仿真结果表明,在信噪比为40 dB的情况下,信号的平均干扰比可以保持在-15 dB以下.因而采用分段平稳和非正交对角化算法相结合的思想来处理信噪比较高时的非平稳信号,可以取得较好的分离效果.

Blind separation of non-stationary signal based on non-orthogonal diagonalization algorithm

Non-stationary signal was processed by piecewise stationary diagonalization algorithm. The computed correlation matrixes of the signal in every subinterval were used to make up a matrix muster. Then Yeredor's non-orthogonal diagonalization algorithm (alternating columns diagonal centers, ACDC) was used in the muster to get the reverse of the separate matrix. So the source signal can be separated from the mixed signal. The simulation results show that ISR (average interference to signal ratio) can be kept below -15 dB when SNR (signal-to-noise) is 40 dB. So when piecewise stationary diagonalization algorithm and non-orthogonal diagonalization algorithm are combined to process non-stationary signal with high SNR,the separation result is very good.