Biocybernetic metrics for workload classification
Abstract
Mental workload is difficult to describe quantitatively. Levels of loading can be compared using the task description, subjective evaluation, or some measure derived from behavioral or physiological responses. Six subjects were tested using a double stimulation paradigm which used up to four lights on a panel to present the stimuli. The paradigm allowed the loading to be varied by changing the inter-stimulus interval (ISI) or the number of valid lights that the stimulus could be selected from (2 or 4 choice loading). Two different inter stimulus intervals (62.5ms and 250ms) as well as a baseline (single stimulation) were used with both choice conditions. Response times and accuracy were measured to provide a measure of the behavioral response. The heart and respiratory rate were measured as an event series only. Four channels of electroencephalographic data (EEG) were recorded to measure the event related potential. A SQUID (Superconducting Quantum Interference Device) magnetometer was used to record the magnetic fields (MEG) produced by the brain. Unfortunately the MEG data was highly corrupted by movement of the head during the task and it was not possible to obtain useful measures from it. Significant differences in behavioral and physiological measures of response were observed for the higher loading levels only. It is believed that the differences between the lower loading levels were not sufficient to cause consistent changes in the measured variables. Response times increased and response accuracies decreased as the loading was increased. Mean heart and respiratory rate did not appear correlated with workload. Increased loading also caused the heart rate variability, as measured by the power spectrum of the inter-beat interval, to decrease. This decrease was not restricted to any particular band of the spectrum. The power spectrum of the ongoing EEG process, as measured from the prestimulus interval, did not yield any useful measures of workload. The latency of the P300 component of the EEG (positive peak nominally located 300 ms after stimulation) was found to increase with loading. P300 amplitude was correlated with stimulus modality, but not workload. It was found to decrease as the number of choices was increased, but increase when the ISI was shortened.
Degree
Ph.D.
Advisors
McGillem, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.