Human Activity Recognition Based on Wearable Devices and Feedforward Neural Networks
James Parluhutan Hutabarat, Nur Ahmadi, Trio Adiono · 2023
Human activity recognition (HAR) using wearable sensors has garnered significant attention in recent years. This technology can continuously track human activities which is essential for remote health monitoring, elderly care, and rehabilitation. In this study, we propose a simple feedforward neural network (FFNN) model equipped with four fully connected layers to enhance the accuracy of HAR. Experimental results conducted on the UCI-HAR (UCI Human Activity Recognition) dataset show that the proposed FFNN model achieve an accuracy and F1 score of 0.96. It outperforms other existing algorithms, which are K-Nearest Neighbors (KNN) and Support Vector Machines (SVM), highlighting the effectiveness of a simplified neural network architecture in enhancing HAR performance.