Human Activity and Posture Classification using Smartphone Sensors and Matlab Mobile

Human Activity Recognition (HAR) is significant, especially in the medical field. Activity recognition has been used in various ways as technology has advanced, particularly using a smartphone-based approach. This work aims to evaluate the accuracy of the triaxial accelerometer in the Matlab Mobile and examine the development and performance of the algorithms in identifying human motions on individuals of similar ages and physical appearances. Motion signals from three subjects are measured, data is preprocessed using a filtering technique, features are extracted, feature normalization is used to reduce bias in data measurement, and activities are classified. Confusion matrix, precision, recall, accuracy, F1-score, and Kappa score are performance indicators used to determine this classification approach. As a result, this research discovered that the Quadratic Support Vector Machine (SVM) produces the best results, with a 99.22 % accuracy rate, proving the efficacy of its activity identification method.