Patterns in Human Activity Recognition Through Machine Learning Analysis Towards 6G Applications

Human activity recognition uses smartphone sensors to identify individual activities. The ability of current sensor-based human activity detection, such as smartphone-based activity recognition, to reliably identify physical activities has declined. Sixth-generation (6G) systems aim to integrate communication and sensing seamlessly. Human activity recognition is a sensing task with many applications in smart homes, emergency systems, and games. It can replace dedicated sensors by monitoring communication signals. Human activity recognition is crucial in several domains and practical applications. Thus, there is a high demand for a support system that can offer information on a user’s present activity by concealing the complex activity recognition. This paper proposed several machine learning methods, such as decision trees, random forests, logistic regression, and K-nearest neighbours, to classify human activity recognition. The proposed methods evaluate the performance of human activity recognition using the WISDM dataset based on the acceleration and angular velocity of a smartphone and smartwatch. The proposed methods accurately identified 18 activities, including walking, jogging, stairs, sitting, standing, typing, brushing teeth, eating soup, eating chips, eating pasta, drinking from a cup, eating sandwich, kicking a soccer ball, playing catch tennis ball, dribbling basketball, writing, clapping, and folding clothes. A comparative analysis was conducted for each activity to determine the precision, recall, and F1 scores. The experimental results showed that K-NN achieved better accuracy at 76.48%, compared to a decision tree, random forest, and logistic regression. A comparative analysis is also performed with state-of-the-art methods.