This research presents a human activity recognition methodology utilizing histogram of oriented gradient pattern history (HOGPH) features and a multi-class support vector machine (SVM) classifier to analyze video streams of activities like browsing, reading, and writing. The system aims to accurately recognize these activities in unconstrained environments, addressing intra-class variability challenges. Experimental results demonstrate that the proposed approach effectively recognizes activities with a reported accuracy of 69.38% using the SVM classifier.