Model for recognizing human behavior via feature and classifier selection

Abstract

Motion or inertial sensors, like the accelerometer and gyroscope that are frequently found in smartphones and smartwatches, can measure the acceleration and angular velocity of bodily movements and be used to teach bots so that they may guess human activities. These models can be selected to a variety of fields, including biometrics and remote patient health monitoring. Because deep learning-based methods employ representing teaching methods, those may automatically identify hidden patterns in data and generate optimal objects from basic information generated from sensors without human intervention, they got popular in earlier in recognizing human activities. Along with recognize human activity, this paper suggests a novel called HDDN-model called CNN-GRU, which combines convolutional units. This model exhibited accuracy that is suggestively better than other state-of-the-art DNN models like Inception Time and Deep Conv LSTM developed using Auto ML, and was successfully verified on the WISDM dataset.