Underdetermined Blind Source Separation Mixing Matrix Estimation And Source Signal Recovery Method Research

Blind source separation has been a hot topic in the field of signal processing, since it has been proposed. Underdetermined blind source separation is blind source separation model, which means the number of the observed signals are less than the number of source signals, this model is more realistic in the field of application. At present, the basis for solving underdetermined blind source separation problem is the sparse component analysis theory, firstly estimate the mixing matrix, then reconstruct the source signal based on it, the premise of this algorithm requires good sparsity of the signals. This paper conducts deep analysis in current research, when the sparsity of the source signals is weak, we study the restoration and reconstruction of the source signal in the case of the underdetermined blind source separation.Firstly, the article describes the underdetermined blind source separation and the theory and model of the blind source separation, and does a major detail on several mainstream algorithms, mixing matrix estimation based on sparse component analysis and major recovery algorithm of the source signal.Secondly, this article gives a brief introduction to the sparse representation and decomposition of the signal, implements sparse representation of the signal by finding sparse representation domain. The signal is converted to the frequency domain by STFT,the data of the signal is processed by the single-source geometric factor in order to make the signal sparse enough. Because the traditional clu stering algorithms need to set the number of classification in advance, this article proposes the number of source signal estimation based on sort cluster of density objects,makes use of the FCM to estimate mixing matrix.Again, in support of the compressed sensing theory, this article derives underdetermined blind source separation model which is equal to the former model, the compressed sensing signal reconstruction algorithm is applied to UBSS signal recovery. In the premise of the mixing matrix can be estimated, this paper trains dictionary and solves signal sparse coefficient, reconstructs the source signal. In this paper, the dictionary training algorithm is K-SVD, and we use OMP algorithm for sparse coefficients.Finally, the underdetermined blind source separation is converted to definite blind source separation problem, EEMD decomposition technique is used for the decomposition and reconstruction of the observed signal, EEMD can Complete the observation matrix of dimension L, the use of more mature Fast ICA algorithm can separate and recover the source signals directly. This method reduces the signal sparsity requirements while being able to recover the source signals.