Freezing of Gait Detection: Deep Learning Approach

Abstract

Freezing of gait (FoG) is one of the Parkinson’s disease (PD) symptoms that appears as an episodic incapability to walk. It usually occurs in patients with advanced PD, and it is a common reason for falls and injury in Parkinson’s disease patients. Freezing of gait must be carefully monitored because it not only decreases the patient’s quality of life, but also significantly rises the risk of injury. In this work, we presented an automatic freezing of gait detection system that is based on the convolutional neural networks (CNNs). The proposed system can perform automatic feature learning and distinguish between freezing events and normal gait. The proposed system eliminates the need for manually extract features and feature selection. The data was collected using five sensors: two telemeters, two accelerometers, and one goniometer. The proposed architecture discriminated the freezing events from the normal walking with an accuracy, specificity, and sensitivity more than 95%.