System for Gait Patterns Analysis Based on Inertial Sensors and Electromyography

Abstract

According to the World Health Organization, more than 1000 million people live with some mobility impairment, about 15% of the global population. Considering the prevalence of people who suffer from mobility impairment, the study of gait patterns is essential to monitor the rehabilitation process, assess emerging technologies for gait assistance and rehabilitation, and study the surgical process’s impact on patients. There are currently some technologies to acquire gait patterns with high robustness (i.e., optoelectronic cameras). However, this mentioned technology is expensive, requires experts for its use, and its protocols are complex to carry out. This master thesis developed a system for gait analysis based on inertial and electromyography sensors, which can be implemented in non-laboratory environments. Stages to develop this project are the following: 1) The validation of algorithms for joint angles acquisition, 2) Developing a method for joint angles and surface electromyography (sEMG) processing, 3) Acquisition of a database of normality for joint angles and sEMG patterns during gait, 4) Implementation of the method for sEMG analysis in a clinical context, 5) Developing an interface to acquire and process joint angles and sEMG patterns, and 6) Evaluation of the system on no controlled environments. Validation of the algorithm to acquire joint angles was performed with a healthy subject, using the gold standard in motion capture: optoelectronic cameras Vicon system (Vicon, USA). Results showed correlation coefficients of 0.97 for knee angle, 0.67 for ankle angle, and 0.95 for the sagittal plane’s hip angle. Besides, relative errors obtained in the range of motion are 3.38% for knee angle, 4.04% for ankle angle, and 10.3% for the hip angle. Additionally, the algorithm was evaluated with eight healthy subjects, in which results show that joint angles follow the normal gait patterns. Regarding the analysis of the electrical muscle activity, an algorithm in the time and time-frequency domain to extract relevant features was implemented in a group of 8 stroke survivors. General results found significant differences in the time and time-frequency features and the waveform of linear envelopes during the gait cycle, compared to people without pathologies. Results show the presented system’s capability to extract relevant information for gait analysis and detect abnormalities related to the joint range of motion and muscle electrical activity. Moreover, the system presents an alternative to the gold standard for motion analysis, in which gait analysis in clinical and in home contexts are suitable to carry out.