Blind Multi-input Multi-output System Identification Based On Second-order Statistics

Blind channel identification aims at retrieving the unknown response of system using only the observed output and the statistics of sources and the characteristics of channels. MIMO system names multi-input multi-output system, its advantages over the traditional SISO system is its large channel capacity. And the blind method based on second-order statistics (SOS) shows the superiority to the method based on higher statistics (HOS) in computations and convergence speeds. Nowadays, Blind MIMO channel identification has widely applied in communications, speech processing and earthquake signal processing, etc. In this paper, we mainly study algorithms of Blind channel identification based on second order statistics in MIMO communication systems.Firstly, the paper reviews the classes and development of blind signal processing, and introduces at large two classes of blind channel identification methods based the received data SOS statistics: Linear prediction method class and Subspace method class, and gives the simulation results and analysis of the two kinds of methods. The merit of Linear prediction method class is that the algorithms are robust to channel order overestimation, we mainly introduce two methods: Linear prediction (LP) and Outer-Product method (OPD). And in subspace method class, we introduce the standard subspace (SS) and minimum noise subspace method (MNS), the advantages of this class are small computations and immune to noise.Secondly, A QR decomposition method (QRD) based on second-order statistics is applied to the blind identification of MIMO systems. This QRD method substitutes the QR decomposition for the single value decomposition (SVD), and decreases the computations of algorithm. Simulations demonstrate that when processing in short data length ORD method has good performance than SS method. In this paper, we then proved the algorithm to be valid for a more generalized case, that is, the source signals are spatial related.Afterwards, we propose a new blind MIMO channel estimation scheme based on signal subspace (SigSS), familiar to BIDS and MNS, the blind MIMO channel estimation scheme divides the MIMO signal space into several identifiable sub-MIMO systems, and then estimates each sub-channel and emerges the all sub-channel parameters into the whole channel parameters. Here, in the signal subspace scheme, a blind channel estimation based on linear prediction is proposed in the paper, we give the simulation results, performance analysis and computations of the algorithm.At last, we sum up blind channel identification methods based on second-order statistics in MIMO-FIR system, and give properly discuss, analysis and outlook.