THE EFFECTS OF LOSSY EEG COMPRESSION ON ERP ANALYSIS

Abstract

This paper analyzes lossy data compression in the specific context of event-related potential (ERP) analysis of electroencephalography (EEG) data. The lossy data compression techniques analyzed here are bit-rate quantization and frequency truncation using the discrete cosine transform (DCT). Within the context of both methods it is demonstrated that ERP analysis waveforms yield significant data compression advantages over raw EEG data. It is found from the experimental results that for any given quantization error bound, utilization of ERP analysis requires approximately 3 fewer bits per EEG sample than normalized EEG data. Additionally, given any error bound for frequency truncation, at least 30% more total DCT coefficients can be discarded when utilizing ERP analysis instead of raw EEG data. The results hold significant implications for large-scale medical applications that rely on ERP analysis of EEG data.