How Low Can You Go? Empirically Assessing Minimum Usable DAQ Performance for Highly Fieldable EEG Systems

摘要

Electroencephalography (EEG) as a physiological assessment technique holds high promise for on-line monitoring of cognitive states. Examples include detecting when a user is overly fatigued, if they are paying attention to a target item, or even detecting sub-conscious object recognition, all of which can be used for greatly enhanced human-system interaction. However, because EEG involves measuring extremely small voltage fluctuations (microvolts) against a potential background that is very large (milivolts), conventional EEG data acquisition (DAQ) systems utilize very high-resolution components, such as low-noise amplifiers and 24-bit sigma-delta analog-to-digital converters (ADCs) on the ideal premise of acquiring a maximal resolution signal to guarantee information content from the data. Unfortunately this comes at the cost of high power consumption and requires expensive system components. We hypothesize that, for many targeted research applications, this level of resolution may not be necessary, and that by intelligently allowing a reduction in the signal fidelity, substantial savings in cost and power consumption can be obtained. To date though a pragmatic minimum resolution remains unexplored. Here, we discuss the utility of using a parametric approach of simulating signal degradation analogous to decreasing ADC bit (vertical) resolution and amplifier fidelity. Results derived from classification of both drowsiness (alpha oscillation) and oddity (P300) detection show strong overall robustness to poor-quality signals, such that classifier performance remains unaffected until resolution is well outside of typical recording specifications. These observations suggest that researchers and system designers should carefully consider that resolution trade-offs for power and cost are entirely reasonable for targeted applications, enabling feasibility of ultra-low power or highly fieldable data collection systems in the near future.