Human activity recognition based on IOT using capsule netwoks and extreme gated recurrnet neural network

Abstract

Human Activity Recognition (HAR) systems have made newlinesubstantial strides in recent years with the help of interconnected sensing newlinedevices and intelligent technologies like Artificial Intelligence (AI), Internet newlineof Things (IoT), and sensors. Sensor-equipped wearable IoT devices are newlineessential for observing and identifying human body movements. Human newlineactivity recognition (HAR) is the process of identifying an activity newlineperformed by one individual or a group of people using spatial and temporal newlineinformation. Applications for computer vision are numerous and require a newlineHAR solution. These include smart home support, medical and healthcare newlineservice monitoring software, and security cameras. newlineRecently, IoT is merged with Machine Learning (ML) and Deep newlineLearning (DL) algorithms to identify human body activities automatically. newlineEarlier approaches showed that the deployment of HAR systems with the newlinehelp of conventional Machine Learning techniques such as Support Vector newlineMachines (SVM), Artificial Neural Networks (ANN), Random Forest (RF), newlineConvolutional Neural Networks (CNN), Recurrent neural Networks (RNN), newlineand Long Short-Term Memory (LSTM). newlineDL algorithms are thought to be more appropriate than ML newlinealgorithms for the purpose of identifying human activities because the HAR newlinesystem comprises of both simple and complicated actions. Motion sensors, newlinesuch as accelerometers and gyroscopes, can be coupled with newlinemicrocontrollers to capture all inputs and send them to the cloud networks newlineusing IoT transceivers in order to monitor human activities effectively. newline newline