Research Article
Abnormal Signal Recognition Method of Wearable Sensor Based on Machine Learning
@INPROCEEDINGS{10.1007/978-3-031-33545-7_23, author={Chao Li and Xuan Zhang}, title={Abnormal Signal Recognition Method of Wearable Sensor Based on Machine Learning}, proceedings={IoT and Big Data Technologies for Health Care. Third EAI International Conference, IoTCare 2022, Virtual Event, December 12-13, 2022, Proceedings}, proceedings_a={IOTCARE}, year={2023}, month={5}, keywords={Machine Learning Wearable Sensor Abnormal Signal Signal Identification}, doi={10.1007/978-3-031-33545-7_23} }- Chao Li
Xuan Zhang
Year: 2023
Abnormal Signal Recognition Method of Wearable Sensor Based on Machine Learning
IOTCARE
Springer
DOI: 10.1007/978-3-031-33545-7_23
Abstract
The recognition of abnormal signal of wearable sensor is of great significance to the application value of the device. In order to improve the accuracy of abnormal signal recognition of wearable sensors and indirectly ensure the safety of wearable sensor devices, a method of abnormal signal recognition of wearable sensors based on machine learning was proposed. According to the different abnormal types and principles of wearable sensors, the signal abnormal judgment criteria are set. The wearable sensor signal is collected, and the initial signal is preprocessed by Kalman filtering, normalization and weighted fusion. The machine learning algorithm is used to extract the features of sensor signals, and the recognition results of the abnormal type, abnormal semaphore and abnormal location of sensor signals are obtained through feature matching. Through the identification performance test experiment, it is obtained that the average abnormal type error detection rate of the optimization design identification method is 0.86%, and the average statistical error of abnormal semaphore is 0.22 db, lower than the preset value.
