Investigation of Temporal Regularities of Accelerometer Signals of Patients Suffering from Gait Freezing in Parkinson's Disease

In this paper, the temporal properties of 3D-accelerometer signals were investigated. The signals were processed using the method of wave train electrical activity analysis. It is based on the analysis of wave train parameters. The wave train is localized increases in the signal's power spectral density over space and time. Using AUC (Area Under ROC Curve), regions in the parameter space were identified that contain differences in the number of wave trains in files with freezing of gait and without freezing of gait. It was found that AUC changes 10–20 seconds before the onset of the freezing of gait. This fact allows us to predict the freezing of gait.