Article Preview
TopConstruction Of The Bsann Based On Improved Hms Model
In the 1980s, ANNs became a hotspot in the artificial intelligence sector due to their unique nonlinear adaptive information processing capability in overcoming traditional shortcomings of artificial intelligence (AI) (Yao, 1999; Mocanu et al., 2018). Neural networks such as fully-connected ANNs have shown excellent performance in classification and regression tasks. However, the prediction accuracy of ANNs is closely related to the model’s complexity due to their particular structure (Hahn, 2007; Muraoka et al., 2005; Rouhani et al., 2010b). At the same time, most existing ANN models for human joint moment prediction use only one fully-connected ANN without considering the possible adaptations of the model structure to the domain knowledge of the joint moment. To compensate for this drawback, the improved HMS model is used to construct a block-wised and sparsely-connected ANN model for the joint moment prediction in this paper.
Hill combined the physiological parameters with the Hill-type model (Buchanan et al., 2004). He thought that the muscle-tendon force can cause the joint moment. This can be written as shown in Equation 1.
(1)In Equation 1,, and are the muscle-tendon’s activation, muscle-tendon’s length, and muscle-tendon’s velocity, respectively. The , , and are the maximal isometric muscle force, optimal fiber length, and tendon slack length, respectively. is the joint angle at optimal length.
The function involves complex and highly nonlinear relations, and the muscle-tendon force must be solved. Another form of it can more intuitively describe the actually calculated muscle-tendon force, as shown in Equation 2 (Xiong et al., 2020).
(2)In Equation 2, and are the muscle-tendon force, is the muscle spans joint angle, and is the time. and are the contractile element and parallel elastic component muscle force. is the pennation angle of the muscle fiber relative to the tendon.