Online Pathological Tremor Characterization Using Extended Kalman Filtering
Résumé
This paper describes different algorithms that perform online pathological tremor characterization. Two distinct nonstationary parametric models are used, an Auto-Regressive (AR) model and an harmonic model. The models are recursively estimated with Extended Kalman Filters (EKFs). Experimental data was obtained with low cost accelerometers and the results are compared in terms of spectrogram estimation and prediction performance.