The wearable sensor system developed by our group, which measured lower limb angles using Kalman-filtering-based method, was suggested to be useful in evaluation of gait function for rehabilitation support. However, it was expected to reduce variations of measurement errors. In this paper, a variable-Kalman-gain method based on angle error that was calculated from acceleration signals was proposed to improve measurement accuracy. The proposed method was tested comparing to fixed-gain Kalman filter and a variable-Kalman-gain method that was based on acceleration magnitude used in previous studies. First, in angle measurement in treadmill walking, the proposed method measured lower limb angles with the highest measurement accuracy and improved significantly foot inclination angle measurement, while it improved slightly shank and thigh inclination angles. The variable-gain method based on acceleration magnitude was not effective for our Kalman filter system. Then, in angle measurement of a rigid body model, it was shown that the proposed method had measurement accuracy similar to or higher than results seen in other studies that used markers of camera-based motion measurement system fixing on a rigid plate together with a sensor or on the sensor directly. The proposed method was found to be effective in angle measurement with inertial sensors.

In this study, development of wearable motion measurement system using inertial sensors has been focused with the aim of rehabilitation support. For measurement of lower limb joint angles with inertial sensors, Kalman-filtering-based angle measurement method was developed. However, it was required to reduce variation of measurement errors that depended on movement speeds or subjects. In this report, variable-gain Kalman filter based on the difference between the estimated angle by the Kalman filter and the angle calculated from acceleration signals was tested. From angle measurement during treadmill walking with healthy subjects, it was shown that measurement accuracy of the foot inclination angle was significantly improved with the proposed method compared to the method of fixed parameter value.

Recently, the use of wearable inertial sensors have been widely studied in the field of human movement analysis. Our research group developed a wearable motion measurement system using the wireless inertial sensors for rehabilitation training and daily exercise. However, there are few reference data to evaluate motor function. In this paper, reference data of joint and inclination angles of lower limb and that of gait event timing for gait evaluation were made by measurement with 4 healthy subjects in their twenties. Average values of inclination and joint angles and gait event timings were similar to those seen in literature. These suggest that the averaged data obtained in this paper can be used as reference data. Then, gait data of a healthy subject in his thirties were compared with the reference data. Most of angles and all the gait event timings were considered to be standard of 20's. However, some angles of the healthy subject in his thirties were considered not to be the standard partly. These differences in evaluation were considered to depend on a level of similarity of movement to the reference data. It was expected to evaluate the level of similarity of movement from various parameters.

No abstract

Abstract-Lower limb joint angle measurement method based on Kalman filter was tested using commercially available inertial sensors in our previous study for a rehabilitation use. Although the angle measurement method was effective, measurement with the wireless sensors were affected by wireless communication environment. Wired and wireless inertial sensor systems were developed in our laboratory for solving the wireless communication problem. This paper aimed to examine the 2 developed sensors, since those were different from the previously used sensors in data communication system or resolution in measurement. The 2 developed sensors and 2 commercially available sensors were evaluated in measurements of inclination and joint angles of a rigid body model comparing to the results of an optical motion measurement system. The angles were measured by correcting the angles calculated from outputs of gyroscopes based on a Kalman filter using outputs of accelerometers. All of the sensors showed small RMS errors under the static and dynamic conditions. The results suggested that various inertial sensors could measure inclination and joint angles stably if angles are calculated by using the Kalman filtering based method. It would be possible to replace sensors with inexpensive or latest commercially available sensors.Index Terms-Lower limb, angle, inertial sensor, kalman filter.