Study On Blind Source Separation Of Communication Signals

With the development of communication technology, modern communication environment has become more complex, the received signals in communication reconnaissance are often a mixture of several signals both in time and frequency domain. As the prior knowledge of the source signals and the mixing system is unknown, the traditional signal processing methods gradually show shortcomings. Blind source separation (BSS) has brought new ideas and methods to separating communication signals.Based on the introduction of BSS-related knowledge, this paper investigates the over determined BSS of communication signals with an unknown source number. First, the widely used natural gradient algorithm is discussed, and the reason to its failing in the over determined situation is analyzed. This problem is solved by improving the algorithm through an orthogonal projection. Simulation results show effectiveness of the improved algorithm.When separating source signals including complex modulation signals (such as OFDM signals), the performance of gradient-based adaptive BSS algorithms deteriorates heavily. In view of this problem and some of its inherent shortcomings, the more practical method based on the joint diagonalization of second order statistics is studied. First, the classic SOBI algorithm is discussed, and some problem of Jacobi orthogonal joint diagonalization adopted in this algorithm is analyzed. Then, improved by a new nonorthogonal joint diagonalization, the performance under a noised environment is enhanced. Many simulation results show effectiveness of the improved algorithm and an evident improvement for performance with low signal to noise ratios.