Hand-arm move control based on Functional electrical stimulation using fuzzy logic systems optimized by reinforcement learning

Electrical signals are naturally present in the human body, with muscles contracting in response to pulses from the central and peripheral nervous systems. Functional electrical stimulation (FES) is a treatment and rehabilitation method where an electric current is applied through the skin to induce muscle contraction. Electrical stimulation near the muscle or nerve can artificially replicate the natural electrical signals, leading to muscle contraction. The electromyogram (EMG) signal, which reflects skeletal muscle activity, correlates with muscle strength and is used to measure bioelectrical events associated with muscle fiber contraction. In this research, arm torque estimation was performed in two stages. The first stage involved estimating the amplitude of the EMG signals, followed by developing a model to illustrate the relationship between the signal amplitude and the torque signal. By employing a fuzzy model, the study aimed not only to reduce estimation error but also to enhance the interpretability and generality of the model. Consequently, the number of fuzzy rules became a critical factor. The deductive clustering method using an ANFIS network, along with another neural fuzzy method that allows for selecting the structure and number of rules, was examined. The efficiency of different methods was evaluated by measuring relative error and VAF percentage, considering the number of fuzzy rules. Sensitivity analysis on the deductive clustering method, using this method with a specific radius, showed a relative error of ۰.۲۴۲ ± ۰.۱۴۵ (mean ± standard deviation) and a VAF percentage of ۹۲.۵۶ ± ۱۲.۱

کلیدواژه ها:

Functional Electrical Stimulation (FES) ، Electromyogram (EMG) Signal ، Fuzzy Logic ، Torque Estimation ، Neuro-Fuzzy Systems ، Reinforcement Learning