Design and data validation of wearable inertial sensor for human gait capture

The article considers the motion capture problem by wearable sensors for mathematical modeling of human gait based on a flat five-link model. Inertial sensors are implemented with a popular accelerometer and gyroscopic sensor MPU-6050, containing a digital signal processor DMP for processing and filtering measurement results. Wireless transmission of wearable sensors data to a personal computer is design using low-power consumption nRF24L01 modules. The optical markers motion capture, which are widely regarded as the “gold standard” for non-invasive motion capture, to sensors data validation was used. The inertial wearable sensors accuracy largely depends of accelerometer and gyroscope data filtering algorithm. During the validation process, three processing methods were tested: a complementary filter, a Kalman filter, and a filter based on DMP. The highest angle measuring accuracy was obtained with DMP-based filtering. The average deviation of hip-mounted inertial sensor data compared to optical markers motion capture was 2.39°, and for the shin-mounted inertial sensor it was 2.42°. The complementary filter based sensor have middle accuracy. A Kalman filter based sensor have the least accurate (a hip angle average error is 5.1° and 7.3° for shin). The developed system of five inertial wearable sensors allows collecting data for human gait modeling using a flat five-link model.