Implementation of continuous wavelet transformation in repetitive finger tapping analysis for patients with PD

Abstract

In this paper we propose a methodology for objective evaluation and classification of repetitive finger tapping performance based solely on its spectral behavior. We used miniature sensor system with gyroscope placed over fingertip of index finger for finger tapping recording. The study included 20 subjects 10 patients with Parkinson's disease (PD) and 10 age and gender matched healthy controls. Acquired data were preprocessed using continuous wavelet transformation (CWT), and their coefficients were used in further analysis. Based on cross-sections of CWT in time and frequency, we introduced parameters describing characteristic tapping frequencies and vigor of the performed movements, its decrement and isolated characteristic frequency areas. These parameters were further used in classification for distinction between PD patients and controls, achieving 95% classification accuracy.