Biometric identification based on human gait provides a non-intrusive and practical alternative to conventional methods that rely on wearable sensors or marker-based systems. Despite the increasing interest in marker less approaches, existing studies often utilize limited datasets, are captured from a single viewpoint, and struggle to distinguish individuals with similar walking styles or physical characteristics. This study introduces a gait-based identification pipeline that utilizes normal walking video data from 17 participants with diverse anthropometric profiles. The system applies Blaze Pose to extract 33 key points from frontal and lateral views, from which 35 features are engineered to represent anthropometric, kinematic, and statistical aspects of gait. These features are then classified using various machine learning models to evaluate performance. The multilayer perceptron (MLP) achieved the best results, with a recognition accuracy of 97.06% and a low false acceptance rate (FAR) of 0.1838%. Compared to prior methods that often focus on constrained data with minimal viewpoint variation, the proposed system demonstrates improved generalizability and robustness when recognizing individuals with similar gait patterns. The use of both anthropometric and dynamic features from multi-angle video further enhances its discriminatory power. Overall, the proposed approach proves to be a reliable and efficient solution for gait-based biometric identification, suitable for real-world applications such as intelligent surveillance and user authentication. It also lays a solid foundation for future studies in more complex or unconstrained environments.