Human Activity Recognition Using 1D Convolutional Neural Network

Abstract

AbstractHumans are more oriented toward innovative research objectives to recognize objects and understand the environment, evaluating time series and forecasting outcomes patterns due to rapid artificial intelligence (AI) advances. Human activity recognition (HAR) is a domain that targets on recognizing, interpreting, and evaluating human movement behavior. HAR has significantly benefitted from deep learning. Despite its immense potential, deep learning models face significant challenges, such as the need for a large dataset for training in the real world. On the other hand, the current study needs to be improved to distinguish static and dynamic behavior with more remarkable achievements. Our main goal is to use one-dimensional convolutional neural network (1D CNN) to create a system that can identify movements such as sitting, standing, walking, sleeping, reading, and tilting, and also we are trying to reduce the time optimization for training the neural network. Human–computer interaction (HCI) mechanization is becoming more in demand for recording behaviors using sensors like gyroscope and accelerometer, and HAR can be done with sensors, images, smartphones, or clips. This paper introduces a technique that employs a 1D CNN to expecting human behavior that is based totally on the dataset given, and we have got finished 90.73% accuracy.KeywordsArtificial intelligenceHuman activity recognitionDeep learningComputer vision1D CNNHuman–computer interaction