Skip to main content

Research Repository

Advanced Search

Adaptive techniques for signal enhancement in the human electroencephalogram

Winski, Richard

Adaptive techniques for signal enhancement in the human electroencephalogram Thumbnail


Authors

Richard Winski



Abstract

This thesis describes an investigation of adaptive noise cancelling applied to human brain evoked potentials (EPs), with particular emphasis on visually evoked responses. The chief morphological features and signal properties of EPs are described. Consideration is given to the amplitude and spectral properties of the underlying spontaneous electroencephalogram and the importance of noise reduction techniques in EP studies is empnasised. A number of methods of enhancing EP waveforms are reviewed in the light of the known limitations of coherent signal averaging. These are shown to oe generally inadequate for enhancing individual EP responses.
The theory of adaptive filters is reviewed with particular reference to adaptive transversal filters usiny the Widrow-Hoff algorithm. The theory of adaptive noise cancelling using correlated reference sources is presented, and new work is described which relates canceller performance to the magnitude-squared coherence function of the input signals. A novel filter structure, the gated adaptive filter, is presented and shown to yield improved cancellation without signal distortion when applied to repetitive transient signals in stationary noise under the condition of fast adaption. The signal processing software available is shown to be inadequate, and a comprehensive Fortran program developed for use on a PDP-11 computer is described.
The properties of human visual evoked potentials and the EEO are investigated in two normal adults using a montage of 7 occipital electrodes. Signal enhancement of EPs is shown to be possible oy adaptive noise cancelling, and improvements in signal to noise in the range 2-10 dB are predicted. A discussion of filter strategies is presented, and a detailed investiyation of adaptive noise cancel liny performed usiny a ranye of typical EP data. Assessment of the results confirms the proposal that substantial improvement in sinyle EP response recoynition is achieved by this technique.

Publicly Available Date Mar 29, 2024

Files




Downloadable Citations