Adaptive Fourier modeling for quantification of tremor

Abstract

A new computational method for quantification of tremor, the weighted frequency Fourier linear combiner (WFLC), is presented. This technique rapidly determines the frequency and amplitude of tremor by adjusting its filter weights according to a gradient search method. It provides continual tracking of frequency and amplitude modulations over the course of a test. By quantifying time-varying characteristics, the WFLC assists in correctly interpreting the results of spectral analysis, particularly for recordings exhibiting multiple spectral peaks. It therefore supplements spectral analysis, providing a more accurate picture of tremor than spectral analysis alone. The method has been incorporated into a desktop tremor measurement system to provide clinically useful analysis of tremor recorded during handwriting and drawing using a digitizing tablet. Simulated data clearly demonstrate tracking of variations in frequency and amplitude. Clinical recordings then show specific examples of quantification of time-varying aspects of tremor.