Adaptive filter for event-based signal extraction

This paper introduces a new adaptive filter structure for a class of event-based signals corrupted by quasi-periodic noise. The event driven adaptive filter emerges from Widrow's adaptive filter, exploiting specific properties of event-based signals such as unpredictable time appearance and limited time duration. This filter can be used when reference input to Widrow's filter, which essentially has to be correlated to the noise, is unavailable, but is known for a fact that it is quasi-periodic. Experimental tests verify improved filter performance of the proposed solution

Duke Scholars

Author Steven A. Cummer Electrical and Computer Engineering