Optimized FPGA Implementation of ICA Based on Negentropy Maximization

Independent component analysis (ICA) is a technique which is used to separate mixed signals. This paper presents an ICA implementation on FPGA utilizing negentropy maximization criteria for updating un-mixing weighting vector. We use this method to separate 4-channel comparatively fast mixed communication signals at the receiver. And before ICA processing, the mixed signals are often required whitening to achieve a better separating performance. We optimized the architectures of whitening and weighting vector updating modules respectively to balance the hardware resource consumption and calculation precision and speed. The performances are evaluated on Xilinx Spartan6 using simulation tool ISim and analysis is presented at the end of this paper.