Detection, Estimation, and Multidimensional Processing of Single Evoked Potentials.
Abstract
A new method of generating and selecting features for computer classification of ERP waveforms is described. The technique employs features that are time samples of measured ERP waveforms that have been bandpass filtered. An improvement of more conventional techniques is shown with both simulated and measured data. A comparison of performance of a quadratic classifier with optimum feature selection is made with one using (suboptimal) forward segmented feature selection. Details of the design and performance are presented for an optimum time-varying filter for estimating ERP waveforms. The filter accepts inputs from multiple electrodes and generates an estimate for the waveform at a single electrode. Results for both simulated and measured data are presented. Some preliminary studies of artifact generation in ERP waveforms are described.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1985
- Accession Number
- ADA170125
Entities
People
- C. D. Mcgillem
- J. I. Aunon
Organizations
- Purdue University