Estratégias de salto em abordagens Markovianas colaborativas para estimativa de ângulos de orientação e articulares

Abstract

This paper proposes the development of an approach for better identification of jumps in collaborative Markovian systems, aiming at a better estimation of orientation and joint angles in consecutive segments of the human body and in articulated robotic devices. The proposal consists of elaborating combined strategies to identify sensors with less dynamic acceleration and at the same time greater relation with the movement, using established strategies of strapdown and finite state machines. For application of the proposal, a Markovian Articular Spatial System was implemented using Strapdown and Gait Cycles. The results obtained were promising, and for the angles of movement in the sagittal plane we obtained a Pearson correlation of 0.80 for the trunk and above 0.996 for the other segments, with an RMSE of 2.17 for the trunk and below 1.43 for others. As for the knee joint angle, a correlation of 0.99 and with RMSE of 1.27 was obtained, which demonstrates the efficiency of the method for segments that have a reliable estimate of orientation, which does not occur with the same accuracy for hip joint, as it depends on the trunk segment.