Data Processing Algorithm And Software Development Based On Epilepsy Monitoring Bracelet

Epilepsy is a common mental disease,the patient base in our country is large,and many new patients every year,this disease has brought great mental pressure to patients and their families,more importantly,is a serious threat to the patient’s life and health safety.At present,EEG signals are mostly used in epilepsy detection.Although this method is mature and accurate,the equipment used to detect epilepsy by EEG signals is relatively large,which is not conducive to portability.Therefore,it is very necessary to develop a wearable device equipped with epilepsy detection algorithm,which is a good news for epilepsy patients and their families.Based on this background,this paper proposes a data processing algorithm based on epilepsy monitoring bracelet,and studies and develops a matching epilepsy monitoring system software.In this project,the device for collecting patient data is the epilepsy bracelet,which is equipped with skin electrical sensor,heart rate sensor and triaxial acceleration sensor to collect these three signals for subsequent analysis.The initial data set is preprocessed,feature extraction and selection,marking and segmentation,and standardization.47 features are extracted,including time-domain features and frequency-domain features,and approximate entropy,sample entropy and fuzzy entropy are introduced.After preprocessing and feature extraction,the data set is input into the machine learning algorithm model.In this paper,the logistic regression algorithm,support vector machine algorithm,Ada Boost algorithm and XGBoost algorithm are compared and studied.Through simulation and comparison experiments,it is concluded that the XGBoost algorithm has the best effect in epilepsy detection,and the accuracy rate of detection of disease is 98%.Due to the limited hardware resources of the embedded device,it cannot support the complex self-learning algorithm.Considering the use of boundary function to monitor epileptic seizures,the test results show that the boundary function of the linear kernel support vector machine algorithm has the best effect,with an accuracy rate of 88.50%,a false alarm rate of 5.14% and a missing alarm rate of 6.36%.In addition,the interpolation method,digital filtering algorithm and window repetition rate used in the preprocessing are compared and the following conclusions are drawn: the third-order B-spline curve interpolation,five-point median filtering algorithm and window repetition rate of 80% are better.In the end,this paper also studies the data flow of epilepsy monitoring system and the interface technology involved in the software,and thus develops the cloud platform software and mobile terminal software.The developed software can run normally,and achieves the expected effect.