Application of optimum linear filter theory to the detection of cortical signals preceding facial movement in cat

Summary

Linear filter analysis was used to detect the occurrence of neuroelectric signals in associated noisy background electrical activity by matching a signal template against incoming neuroelectric data. Bach signal to be detected in the neuroelectric data consisted of a gross potential change recorded at the coronal — precruciate cortex of a cat, evoked by an auditory conditional stimulus, and related to the production of a conditioned facial movement. Detection of the occurrence of the signal corresponded closely to detection of the ensuing movement.

The operation of the matched filter on the signals in noise was studied for different threshold levels of detection. Threshold settings were selected to maximize successful detections and to minimize false alarms. The results of our experimental detections agreed closely with predicted, theoretical detection levels derived from Wiener's models of optimum detection of signals in additive noise. Levels of detection were found to depend upon the signal to noise ratios and frequency spectra of the analyzed data and could be predicted a priori from a knowledge of the latter on the basis of Wiener's theory. The ability to predict optimum detection levels, by the linear filter method, of cortical electrical signals related to the production of movements may provide a basis for evaluating the merit of such signals in the design of prosthetic devices for motor control.