Design And Implementation Of SEMG Prosthetic Controller Based On ARM

Nowadays, the number of the upper limb disabled people in China is increasing.Due to its high cost, large volume and unnatural use, the traditional prosthesis bringthe disabilities inconvenience in their daily lives. With the development oftechnology, a intelligent prosthesis based on surface electromyography (sEMG) hasbecome a hot topic.This bionic arm can achieve the appropriate action according tothe sEMG signals, improve their self-care ability greatly, so as to let them integrateinto socieyt.sEMG signals is a kind of biological signals generated by the muscleactivity，which collected from human epidermis. The sEMG contains a wealth ofinformation about muscle activity, which reflecting the state of human limbmovement. Therefore it can be used as a source to control prosthetic, so as to achievecontrol prosthetics with multi-degree freedom.For the issue that most electromyography prosthesis controller systems existedare on basis of the threshold to control the hand’s opening and closing which bringspoor operational flexibility, a design scheme of electromyography prosthesiscontroller system based on ARM is proposed. The scheme can achieve real-timeonline identification of ifve kinds of action mode, containing the palm up，palmdown，ifst，palm and no action, with only two HMG sensors.The system is divided into hardware and sotfware components. The hardwarecomponent includes system power module, processor minimum system module,signal conditioning modules and serial communication module. The softwarecomponent includes A/D signal acquisition, IIR filtering, signal characteristicsextraction，BP neural network recognition.Firstly, the system collccts sEMG of biceps and triceps by two sensors. Becausethe original sEMG signal is weak and negative, we need to amplify and adjust thissignal, and put it into the STM32ARM core processor to A/D sampling, with thesampling frequency2.4Khz. Secondly, in the processor, the sEMG is beingband-pass filtered by use of the IIR digital filter, and four kinds of time-domain and frequency-domain features are extracted， which consist of the signalamplitude average absolute value (MAV)，the signal standard deviation (SD), medianfrequency (MF) and the mean power rfequency (MPF). Thirdly，a8-dimensionalfeature vector is composed by four features of two sEMG signals. The weight andthreshold values are acquired by training BP neural network, and calculated bySTM32according to the BP network layer to layer transfer function，for results of5kinds of movements. Finally, the prosthesis is being controlled to makecorresponding action according to the output control signal result from theidentification.We has made a large number of simulation and actual experiments. Theemperiments results show that, the recognition rate can reach98%under Matlabsimulation，and97%under the actual online recognition experiments. Therefore, thesystem is able to meet the real-time requirements and the needs of the prosthetichand control.