In many real-world applications, such as sports analytics, smart settings, and healthcare monitoring, Human Activity Recognition (HAR) is indispensable. The objective of this study is to leverage time-domain information taken from the HARTH dataset to enhance HAR performance. In order to improve the precision and effectiveness of activity recognition, the study looks into various machine learning classifiers. The objective of this work is to create a more effective method for differentiating between different physical activities by analyzing important time-domain properties as mean, variance, skewness, kurtosis, and signal magnitude area (SMA). To determine the best classifier for HAR tasks, a number of machine learning models are assessed, including Support Vector Machines (SVM), Random Forest, and Decision Trees. Based to the results, well-designed time-domain features greatly improve classification performance while lowering computing cost and preserving high accuracy.