Research On Sedentary Behavior Recognition Based On Wi-Fi Channel State Information

In today’s rapid development of information technology,Internet of Things(Io T)technology is continuously popularized,and human-computer interaction is a popular direction in the current field.At the same time,sedentary behavior has become a common phenomenon in people’s working life,but people know little about the health risks of sedentary behavior.In this paper,we provide two innovative system solutions for sedentary behavior recognition using Wi-Fi channel state information,using machine learning models and parallel Long Short-Term Memory(LSTM)neural networks and Convolutional Neural Network(CNN).to perform sedentary behavior recognition,the main work of which is as follows.Translated with www.Deep L.com/Translator(free version)1.Design a Wi-Fi sedentary behavior recognition system based on machine learning model.By extracting the channel state information features affected by the actions in the user’s sedentary behavior,the recognition of the user’s sedentary behavior is achieved by borrowing a machine learning classification model.In the whole experiment,firstly,the experimental setup is adjusted to enhance the information of the changes caused by the sedentary behavior on the channel state with the help of Fresnel zone theory and Rice distribution theory.The data are then processed by outlier detection,median filtering,low-pass filtering,etc.to retain the valid information.Finally,the data are sliced and labeled as action information.Subsequently,15 experimental persons with 4 different behavioral data were identified by using different machine learning classifiers for behavior recognition.In the experimental results,the highest accuracy of 93.3% was achieved using the support vector machine classifier.2.Design a deep learning network based Wi-Fi sedentary behavior recognition system.The feature extraction of time series data after principal component analysis and CNN for mapping energy map image data are performed using LSTM network and sedentary behavior recognition after feature fusion,respectively.The experimental results seed that the LSTM+CNN network using the combined two networks has a better recognition rate of up to 95.1%.The effects of different systems,different genders and body types of experimenters on the classification results were also compared to further validate the robustness of the system.