Sistema Markoviano espacial autoalinhável para estimativa angular de atitude e articular entre dois segmentos consecutivos

Abstract

The interest in gait analysis and daily activities can find applications in several branches of human activity. With continuous research and development, motion capture devices are getting smaller and more affordable in terms of cost. This dissertation presents the development of a system for estimating the attitude (inclination and direction) and joint angles between two consecutive lower limb segments of the human body, using only accelerometers and gyroscopes. In order to improve the accuracy of the estimates and compensate for the numerical integration errors of the gyroscope, each sensor system, installed in its respective segment, performed data fusions through a Kalman Markovian filter, operating in a collaborative way between the systems of sensors at times when the accelerometer measurements registered values below a determined threshold. The strategy of Markovian systems was used, whose jumps choose the type of observation performed on the system, which can be: (a) nominal, considering both segments, (b) local, choosing the segment with the lowest dynamic acceleration index. Unlike previous studies that use Markovian systems with inertial sensors and encoders for absolute angular estimation in lower limb exoskeletons, this study extends the use of this type of system in devices where the encoder is not present, or even in kinematic and dynamic studies of the human body.