Research On Automated Movement Pattern Classification Of Lower Limb Based On Surface EMG Signals

The surface electromyography is a kind of comprehensive result on the time and the space of muscle electromyography. It is the manifestation of the neuromuscular activation associated with a contracting muscle. This dissertation mainly studies lower limb surface EMG signals processing and movement pattern recognition methods. It belongs to the cross-subject research of neuro-electromyographic signal of neuromuscular science, signal processing and pattern recognition, etc.With the development of computer technology in recent years, the sEMG signal has attracted more and more scholars’attentions. It has been not only applied to sports medical science, the clinical medical science and healing medical science, but also suggested and utilized as an effective method to provide control commands for artificial limbs and functional neuromuscular stimulations. This dissertation studies the problem of how to identify the different movement pattern of lower limb by the sEMG signal because sEMG signal application bases on its pattern recognition. The purpose is to suggest the theory basis for the application of mult-free sEMG controlling artificial limb and uncover sEMG essence by picking up its effective signal features and to design pattern classifier according to the nonstationary and randomness of the sEMG signals and modern signal processing methods as well as modern pattern recognition theory. The mainly work and conclusions are as follows:1. The sEMG signal filtering circuit is the key of collection system. According to sEMG amplitude-frequency characteristics and the impact of the outside, it designs the better filtering circuits, particularly in the design of a new type of50Hz frequency circuits which can be satisfactorily resolve50Hz frequency noise.2. In the dissertation we classify six different movement of lower limb and complete the estimate of muscle fatigue and different road conditions combing wavelet theory with traditional analysis ways of time domain and frequency domain, on the conditions that taking BP and SVM as the classifiers. 3. The dissertation propose a new process neural network theory based on the PSO optimizing algorithm. Not only it considered the space coupling effects of the sEMG, but also considered the effects of storage time of the signal. It prevents losing information of human’s lower limbs without extracting features of sEMG.. At the same time, compared with traditional PNN, the algorithm we proposed in the dissertation has superior executive efficiency and higher precision rate.