Independent Component Analysis Research On Objective Function And Strategy Optimization

Independent component analysis (ICA) is an important branch of blind source separation (BSS). It recovers source signals from received mixture based on the source signals’ statistical characteristics of independence. ICA becomes more and more important in many areas such as biomedicine, speech and communications, image processing, earth science, data mining and so on. It has a very important practical value and application prospects. However, traditional ICA algorithm has the problem of narrow scope, slow convergence and low separation accuracy. To meet the demand of a wide range of practical applications, algorithm optimization research of ICA becomes attractive and far-reaching.Aiming at the problems above, the following work was done based on the previous research:1. ICA’s theoretical summary research. ICA is mainly composed of objective function and optimization algorithm. Firstly, different objective functions were derived and pointed out similarities; the optimization algorithm was then divided into two categories:numerical method and swam intelligence method, followed by analysis of optimization algorithms’main strengths and weaknesses. The swam intelligence method was decided as the study point. And artificial bee colony (ABC) algorithm was proposed as the optimization algorithm, its bionic strategy was discussed.2. The objective function and optimization strategies of ICA were optimized to improve algorithm. For the objective function, the Givens rotation was used to reduce the amount of calculation; for the optimization strategy, an adaptive strategy was proposed. An adaptive global guidance item was introduced to dynamically adjust optimal guiding role, and adaptive Boltzmann probability was adopted in selection strategy to adjust selective pressures.3. Simulation experiments were conducted for the comparison of different algorithms’ performance. The results showed that, modified algorithm could separate signals of different kurtosis types successfully. Compared with other algorithms, modified algorithm significantly improved the separation accuracy to about three orders of magnitude and.had good performance.