THE EFFECTS OF LOSSY FREQUENCY-DOMAIN EEG COMPRESSION ON CROSS-FREQUENCY COUPLING ANALYSIS

Abstract

This paper analyzes lossy frequency-domain compression in the context of cross-frequency coupling (CFC) analysis of electroencephalograph (EEG) signals. The approach used here for CFC analysis involves a low-complexity signal analysis block followed by a constant false alarm rate (CFAR) detection algorithm. The lossy frequency-domain compression is achieved via the threshold coding method for frequency truncation using the discrete cosine transform (DCT). This method is found to increase CFC detection rates by as much as 30% to 50% depending on the amount of Gaussian noise in the signal and the selected probability of false alarm. Further analysis indicates that these significant improvements in CFC detection rates are due to adaptive frequency-domain noise reduction. These results bode well for lossy frequency-based EEG compression schemes which can greatly improve transmission speeds and decrease storage space requirements while simultaneously enhancing CFC analysis capabilities.