Abstract: Independent Component Analysis (ICA) decomposes observed mixed random vectors into statistically independent components while optimizing the contrast function. This study introduces a new automatic method to improve the ICA algorithm and its Signal to Noise Ratio (SNR). As the convergence speed is based on the contrast function, it is improved by optimizing the contrast function with hardware optimization technique. The proposed ICA technique is validated by retrieving the maternal and fetal Electrocardiogram (ECG) signal from their mixtures. Floating-Point (FP) arithmetic calculations are performed to increase the number precision and dynamic range of the signals and hence SNR. Simulation and synthesis are done using Quartus II tool and the algorithm is implemented in Cyclone IVGX Family FPGA. The proposed algorithm operates at a frequency of 2.21 MHz and it provides mean SNR of 47 dB. The proposed method is also implemented in 0.18 um standard cell CMOS technology using Industry standard tool Cadence.
Jayasanthi Ranjith and Muniraj NJR, 2016. ECG Extraction by Improved Independent Component Analysis. Asian Journal of Information Technology, 15: 2638-2644.