FPGA implementation of optimized independent component analysis processor for biomedical application

Independent component analysis (ICA) is a statistical signal processing technique for separation of mixed voices, images and signal. The basic idea of ICA is to find the underlying independent components in the mixture by searching for a linear or nonlinear transformation and minimizing the statistical dependence between components. Due to the computational complexity of ICA and commonly used data sets, the ICA process is very time-consuming. For reducing the complexity of ICA algorithm, modularity, hierarchy and parallelism are introduced in VLSI implementation. It is more efficient when the cost function, which measures the independence of the components, is optimized. System level design of ICA with evolutionary optimization algorithm is proposed for EEG signal processing. The use of evolutionary computation based optimizations i.e Adaptive Shuffled Frog Leap Optimization Algorithm with additional operations of mutation, crossover and feedback resolves the permutation ambiguity to a large extent [8]. This ensures the convergence of the algorithm to a global optimum and its VLSI implementation ensures high speed processing.