Automated Gait Analysis using a Kinect Camera and Wavelets

Studies of Parkinson’s Disease (PD) have generated particular interest in factors such as gait and posture patterns and fall risk. Human gait patterns involve a basic gait cycle that is composed of two phases: stance and swing. With gait analysis, we can derive values for spatiotemporal variables such as the walking speed, cadence, and stride length from these stance and swing phases. In this paper, we use a low-cost, quick-setup, and portable system to capture gait signals, and propose a novel method for automatically obtaining gait phases (swing and stance) using wavelets and a Kinect camera. We tested this method on six PD patients and six healthy subjects in a clinical context, finding that it could classify the gait phases with 93% accuracy, compared with clinical judgment. Such a procedure could allow clinicians to rapidly, easily, and non-invasively diagnose and assess PD patients via objective and automatic data analysis.