A Visual Calibration Driven Gait Analysis Model using Wearables

Gait analysis is a critical tool in clinical research, providing invaluable insights into the locomotor abilities and potential underlying neurological or musculoskeletal conditions of patients, particularly those afflicted with stroke and athetosis. Currently, in gait analysis research, there is a predominant reliance on either single wearable or non-contact sensors, resulting in issues such as a lack of comprehensiveness in gait parameter analysis and the absence of control for calibration. Nowadays, a system combining both sensors has become a common approach to address these challenges above. However, while numerous researchers have proposed a variety of such systems, the majority of them need more adaptability and universality. Hence, we proposed an adaptive general gait analysis model based on IMU data calibrating by vision information (AdaGA-VI). In this model, regardless of the placement of sensors on the body, the positional changes between each sensor can be accurately calculated using quaternions. Experimental results indicate that our model enhances the accuracy of gait parameter analysis and exhibits extensive applicability to other analyses involving quaternions.