No abstract

Electrogoniometers are low price sensors which are easily attached to the body in any environment. Although electrogoniometer sensors are typically used for angular measurements, they can also be used to determine position. This study aimed to accurately determine hand position during the performance of five daily life activities using two electrogoniometer sensors to measure shoulder and elbow angles simultaneously. The measurement of joint angles involves some errors which are divided into intrinsic and extrinsic errors. These errors cause considerable inaccuracies in the estimated hand positions. To overcome this issue, the errors identified are compensated for in two phases, the angular phase and the positional phase, in which a polynomial function and an Elman neural network are used for error compensation, respectively. The derived hand trajectories and the decrease of the root mean square error at every stage of the error compensation, along with the variance accounted for of the Elman networks for each task, show the effectiveness of the proposed method.