The goal of the project is to identify a person-of-interest based on specific set of soft biometric attributes from a surveillance video. The project will integrate and minimize semantic gap between human descriptions and soft biometric traits. Moreover this project aims at extracting soft biometric features from an input image or a video frame and then uses these features to identify the person. This project will add another dimension to surveillance as human cognitive perceptions are used while searching the video streams.