Developing a new foot muscle model of gait using a Bayesian network

In this paper, by means of a statistical method, we use sole pressure on each part, angle of joint, femoral and crural muscle activities to produce a new foot muscle activity model for use in the design of ankle-foot orthoses. We built a Bayesian network model[1] by examining the normal gait of a nondisabled subject. We measured the activity of the lower foot muscles using electromyography, joint angles and the pressure on different parts of the sole. From these data, we built three models, representing the stance phase, the control phase and the propulsive phase. The accuracy of these models was confirmed. The largest feature of this model is making every 10% level nodes of each measurement data. Normal Bayesian network can estimate only muscle active or not active. But this method can estimate activity level of muscle. From this feature this method has three advantages. First, our use of 10% increments in the levels of the measured factors enabled changes in these factors during gait to be reflected in the model. Second, variations in the influence of factors that differ between low and high muscle activity are represented. Third, it is easier to use than physical models; three-dimensional motion analysis is not required and the method is convenient for clinical use. In an evaluation of this model, we confirmed that this model can estimate all muscular activity level with an accuracy rate greater than 95%.