Window length effect on cross frequency coupling in an EEG processing circuit

This paper examines the window size selection used for measuring cross frequency coupling (CFC) in digitized electroencephalograph (EEG) signals. As with many signal processing algorithms, a larger time window results in more accurate and less noisy results but at the cost of a lower temporal resolution. In this paper, a range of window sizes (in samples per second) is examined using a synthetic test signal and, based on this analysis, the window size selection is made for a hardware implementation of the modulation index CFC algorithm. We found the optimized window length is 512 samples/second for EEG signals. An ASIC has been created from this design using the IBM 0.18 micron process with a chip size of 3mm × 3mm with 191mW power consumption and is currently being tested.