Research And Implementation Of Gait Detection Algorithm Based On Plantar Pressure

Hallux valgus is a common foot disease.Excessive foot pressure in patients with hallux valgus could causes various pains,which would in turn damages their knees and ankle joints.If not treated in time,it will lead to joint disease and its secondary pain might seriously affect people’s walking and running functions and change the human gait.The change of plantar pressure can reveal the dynamic characteristics of the foot during human movement,which is an important basis for reflecting whether the human gait is normal or not.In clinical medicine,gait analysis can provide objective evaluation for the diagnosis of various foot and ankle diseases,therefore the gait detection which is based on plantar pressure has important research value.Based on the pressure information of the foot,this paper combines with the theoretical basis of the gait detection classification algorithm and utilizes the convolutional neural network optimization algorithm,which completes the realization of the hallux valgus disease identification algorithm.And this paper provides a better solution with lower false positive rate for the hallux valgus disease recognition.The main research contents of this paper are as follows:(1)With the human gait features studied,the plantar pressure characteristics of hallux valgus are analyzed in detail,and a classification model for hallux valgus gait detection based on plantar pressure information is designed.A gait detection algorithm framework based on convolutional neural network is proposed;A feature extraction network and a classification network are constructed,and a gait detection classifier model is constructed to implement hallux valgus gait detection based on plantar pressure.(2)As for the problems of too many plantar pressure data frames and redundant information in a single sample,an algorithm for extracting key frames of data is proposed.This algorithm can automatically filter out data frames with complete plantar pressure information to reduce the amount of parameter calculation in the training process and remove redundant data.Through the design of comparative experiments,the accuracy of gait detection classifier model is 88.33%,which verifies the effectiveness of the algorithm.(3)Aiming at the problem of sample imbalance in the data set,the SVM algorithm is studied,and introducing different misclassification costs of each category,the SVM is improved into a cost-sensitive support vector machine(CS-SVM).The improved CS-SVM algorithm is suitable for the case where the sample ratio of hallux valgus patients and healthy people is 1:3,which emphasizes the importance of minority sample data in the classification task and effectively reduces the loss caused by misclassification.The gait detection experiment is designed and verified.The results show that the improved classifier model can identify hallux valgus.The classification accuracy,sensitivity and specificity of the classifier are 93.33%,90.4%and 96.52%,which shows that the algorithm in this paper has better classification performance and can effectively reduce the misjudgment.