Research On Multi-systems Adaptive Combination Based Blind Source Separation Algorithm With Momentum

The process of separating the source signal based only on the observed signal is called blind source separation.Blind source separation technology is widely used in many disciplines such as signal processing and neural networks due to its unique properties.Based on the adaptive combination of multiple systems,this paper focuses on the research of the natural gradient with momentum term based on blind source separation algorithm,and has made many improvements and complete to the existing algorithm.The main research work of the full text includes the following:First,taking the background and significance of blind source separation research as an entry point,the research history and current status of blind source separation are given,and the knowledge of mathematics and information theory involved in blind source separation is outlined in detail.At the same time,how to optimize the blind source separation algorithm and the criteria for judging the separation performance are briefly introduced,which provides basic knowledge for the theoretical proof and simulation experiments of the next algorithm.Secondly,according to different signal processing methods,the implementation principles of several commonly used blind source separation algorithms are introduced.The adaptive blind source separation algorithms are mainly analyzed,including natural gradient algorithm,EASI algorithm,and so on.The advantages and disadvantages of the algorithm are discussed.Then,based on the natural gradient blind source separation algorithm,the focus is on the optimization of the algorithm's convergence speed and steady-state error.The main innovations are as follows: 1.The momentum factor is integrated into the cost function,and the iteration result of the natural gradient with momentum term based on blind source separation algorithm is derived using natural gradients,which makes up for the lack of theoretical basis for the existing natural gradient with momentum term based on blind source separation algorithm;2.Based on 2-MNG algorithm,considering the constraints on the performance range of traditional combination factors,combined with adaptive momentum factor technology,a new 2v-OMNG algorithm is proposed,and theeffectiveness of the algorithm is verified through simulation experiments;3.The momentum term is used to improve the convergence rate of the algorithm.On the basis,4-MNG algorithm is proposed by combining different step size parameters,and matrix transfer is used to improve the proposed algorithm,which further optimizes the contradiction between algorithm convergence speed and steady-state error.Meanwhile,using simulation experiments confirmed the proposed scheme can improve the overall performance of the adaptive separation algorithm.Finally,the full text is summarized,and the problems still facing in the research of blind source separation technology and the future development direction are prospected.