EEG monitoring: Performance comparison of compressive sensing reconstruction algorithms

Abstract

EEG represents the electrical activity across brain. This activity is monitored to diagnose the diseases due to brain disorders, like epilepsy, coma, sleep disorders, etc. To record EEG signal, a minimum of 21 electrodes are placed across the scalp, which generates huge amount of data. To handle this data, compressive sensing (CS) proves itself to be a better candidate than the traditional sampling. CS generates far fewer samples than that suggested by Nyquist rate and still allows faithful reconstruction. This paper compares the performance of CS reconstruction algorithms in reconstructing the EEG signal back from compressive measurements. The algorithms compared from convex optimization are basis pursuit (BP) and basis pursuit denoising (BPDN) and from greedy algorithms are orthogonal matching pursuit (OMP) and compressive sampling matching pursuit (CoSaMP). The performance of these algorithms is compared on the basis of speed and reconstruction efficiency.