Gait Cycle Analysis Algorithm Based on Depth Camera Measurement

Gait analysis is widely used in the evaluate in medical gait diagnosis, patient rehabilitation training, and scientific training guidance. Gait analysis system effectively reconstruct 3D human joints model to obtain related gait information through specialized calculations. In this paper, we present a Real-Time Gait Analysis Algorithm (RTGA) based on RGB-D images input, which replaces traditional methods that capturing human body motion trail from wearable inertial measurement units (IMU) equipment. Combining real-time gait data acquisition and a deep learning-based gait medical feature analysis algorithm, we propose a gait measurement and acquisition system based on reliable 3D human joints distribution. We first take joint sequence smooth algorithm to filter the noise data the original 3D motion curve. Then, we perform joints transformation and position correction for the proposed method, which eliminates the gait cycle detection error caused by random noise interference. Finally, we propose a well-designed evaluation system to identify gait cycle and calculate each medical index (such as step width, speed, frequency). We employ Azure-Kinect DK to acquire 3D human body semantic content as input and validate the effectiveness of our RTGA system in the wild.