Research On Partial-hand System Design And Pattern Recognition Algorithm Based On FSR Sensor

In patients with upper limb disability,some patients with handicap handicapped account for a large proportion,and half-time studies of this group of people have just started.There is currently a lack of functional prostheses that are suitable for this group of people.In this paper,a highly integrated multi-degree-of-freedom partial hand system based on FSR(Force Sensitive Resistor)pressure sensor is proposed.The pattern recognition algorithm and control experiment are completed by acquiring the muscle strength signal.This paper first introduces common signal sources for half-palm hand control,introduces the application of mechanical driving signals,EEG signals,EMG signals and FSR pressure signals in the control of prosthetic hand,and points out the advantages of FSR sensors.A partial hand design based on FSR pressure sensor is proposed,and an acquisition system and pattern recognition algorithm based on FSR pressure signal are proposed.In order to meet the space and size requirements of a partial hand hand,this paper first proposed a partial hand design that integrated the drive motor into the inside of the finger,and established a muscle force signal acquisition system and a control platform.The muscle force signal acquisition system uses the FSR sensor to measure the contraction and extension of the residual muscles of the palm to obtain the pressure signals corresponding to different hand postures.An 8-channel FSR pressure signal realtime data acquisition system was established.The signal analysis and test platform based on Lab VIEW was used to identify and classify the signals.The STM32 microprocessorbased control system was used to implement the partial hand movement control.Then,a pattern classification algorithm for muscle strength signals is proposed.The two main methods are SVM(Support Vector Machine)and MLP(Multilayer Perceptron),which discuss different kernel functions and different classifications.The effect of the method on the recognition accuracy is optimized using grid search methods.Combining SVM and MLP models is proposed and compared with other methods,and the advantages of combinatorial algorithms are pointed out.Finally,based on the 5-fold cross-validation experiment,the change of recognition rate under repeated wear is obtained,and the effectiveness of the algorithm is proved by experiments.Finally,a partial hand test control platform was built to test the performance of a single finger,and the application of the FSR partial hand hand in actual grasping and coordination movement was completed.For 9 types of gestures commonly used in life,the training and classification experiments based on FSR signals are completed.The success rate reaches 80%,and satisfactory results are achieved.This can meet the daily needs of the disabled and has a good application prospect.