Abstract: Human gait is a behavioural biometric that allows for person recognition from the patterns of walking. In contrast to face or fingerprint recognition, gait recognition can be done at a distance without subject participation, making it particularly valuable for surveillance and access control applications. This paper describes a real-time human recognition system that utilizes gait-based features extracted from Media Pipe pose landmarks, combined with a light K-Nearest Neighbours (KNN) classifier. The system is able to run on generic-purpose hardware utilizing a web-based interface constructed with Flask. Both offline acknowledgment through video upload as well as in real-time through webcam input is supported. Deploy ability is one of the essential strengths of this initiative: it uses no GPUs, big data, or complex training pipelines and yet provides consistency in accuracy as well as response. This places it in line to be considered a top prospect for edge-based intelligent systems across public safety, smart cities, and IoT systems.

Keywords: Gait Recognition, Human Identification, Real-Time Recognition, Pose Estimation, Media Pipe, OpenCV, Flask Web Application, KNN Classifier, Biometric Authentication, Skeleton Tracking

| DOI: 10.17148/IARJSET.2025.12430