Electromyography-assisted optimization (EMGAO) approach is widely used to predict lumbar joint loads under various dynamic and static conditions. However, such approach uses numerous anthropometric, kinematic, kinetic, and electromyographic data in the computation process, and thus makes data collection and processing complicated. This study developed an electromyography-based support vector machine (EMGB_SVM) approach for predicting lumbar spine load during walking with backpack loads. The EMGB_SVM is simple and uses merely the electromyographic data. Anthropometric information of 10 healthy male adults as well as their kinematic, kinetic, and electromyographic data acquired during walking exercises with no-load and with various backpack loads (5%, 10%, 15%, and 20% of their body weight) were used as the inputs of a biomechanical model, which was then used for predicting the lumbosacral joint compression force. The efficacy of the EMGB_SVM was investigated by comparing the force profiles obtained using this model with those obtained using the current EMGAO approach. On average, the EMGB_SVM obtained deviations in the peak and minimum forces of −3.3% and 5.1%, respectively, and a root mean square difference in the force profile of 7.5%. The EMGB_SVM is a comparable estimator in terms of its slight bias, favourable consistency, and efficiency at predicting the lumbosacral joint compression force. Copyright © 2019 IPEM. Published by Elsevier Ltd. All rights reserved.
CitationLi, S. S. W., Chu, C. C. F., & Chow, D. H. K. (2019). EMG-based lumbosacral joint compression force prediction using a support vector machine. Medical Engineering & Physics, 74, 115-120. doi: 10.1016/j.medengphy.2019.09.009
- Machine learning
- Spinal loads
- Trunk muscle