| Muscle fatigue analysis and muscle force prediction have been widely used in many fields, such as sports medicine, rehabilitation medicine, biomechanics, etc. Muscle fatigue analysis can provide useful information for rehabilitation exercises and physical exercises. And muscle force is one of the most important parameters in some research fields including prosthesis control, biomechanics and human interaction. As a non-invasive and effective method to evaluate the activation level and functional status of muscle, surface electromyography (sEMG) plays important role in the research of muscle fatigue and muscle force prediction.Although lots of investigations have been carried out on muscle force prediction in past years and some achievements have been obtained, the applicable conditions of muscle force prediction is very limited due to the complexity of neuromuscular system. Most of the related studies focused upon particular muscles (big muscles with clear functions, such as bicep, tricep and vastus lateralis) with fixed contraction patterns (such as static contraction and cyclical contraction). With the assumption that the muscle was under non-fatigue state, little attention has been given to the influence of fatigue on muscle force prediction.This thesis aimed at exploring the feasibility of muscle force prediction under muscle fatigue. To reach the research goal, the performances and limitations of several sEMG parameters in muscle fatigue analysis were compared firstly, and then a modification method for muscle-force-model was proposed to realize the prediction of muscle forces in both unfatigued and fatigued states, which can enhance the practicability of muscle force prediction. Additionally, a real-time sEMG analysis system was developed based on the research results of muscle fatigue analysis and muscle force prediction.The main work of the thesis can be summarized as follows:(1) Muscle fatigue analysis. The performances and limitations of several sEMG parameters in muscle fatigue analysis were analyzed. These parameters include MPF (mean power frequency), MDF (median spectral frequency), FI (Dimitrov spectral index), ARC1(the first autoregressive model coefficient) and RMS (root mean square). The experimental results showed that MPF outperforms other parameters in muscle fatigue analysis, considering monotonicity, stability and repeatability. Experimental results also showed that the relation between parameters and the level of muscle fatigue could be influenced by muscle contraction pattern, the position of the electrodes and individual differences.(2) Muscle force prediction. Firstly, the mathematical model of muscle force prediction under non-fatigue state was established using polynomial curve fitting (a quadratic multinomial was used as the fitting curve), and experiments were conducted to explore the influence of some factors on the accuracy of prediction. Results showed that the position of electrodes and the muscle shape would significantly influence the prediction results, but the impact of contraction pattern was small. Then, the influence of muscle fatigue was analyzed emphatically, and the prediction error of the muscle-force-model was found to rise with the deepening of muscle fatigue under both constant and cyclical contractions. At last, dynamic modification methods based on muscle fatigue were separately set up for the two patterns of muscle contraction. Experimental results demonstrated that the prediction error of modified muscle-force-model remained stable and low at various fatigue level, which was close to the error under non-fatigue state.(3) Real-time muscle fatigue analysis and muscle force prediction system. A real time sEMG analysis system was developed based on the research results of our study. This system provided flexible supports for data acquisition, signal display, muscle fatigue analysis and muscle force prediction, and it can be used for both real-time and offline analysis. |
PDF Full Text Request