A framework for suspect face retrieval using linguistic descriptions

Abstract

Facial sketch plays a vital role in suspect’s identification and apprehension by law enforcement agencies. The sketch artist develops sketches based on the memory of the eye witness. However, sketch artists are less in number and limited to availability. It is also observed that as time passes, the onlooker even forgets many of the critical attributes, which proves costly in time-sensitive investigations. State-of-the-art sketch-photo retrieval methods have used the sketch to retrieve the suspect’s photograph and ignore the importance of time sensitivity. In this paper, we have proposed a linguistic description based approach for suspect’s photo retrieval. In the proposed approach, the facial attributes and their descriptions are extracted from the input linguistic information using parts of speech and attributes ontology. Attributes are ranked according to their importance in the identification of a suspect. An ensemble classifier which aggregates the predictions of multiple classifiers is used to retrieve a suspect’s image. Experiments are performed on standard datasets. The performance of the proposed approach is evaluated by comparing it with the existing methods of linguistic sketch-based retrieval as well as with sketch to photo retrieval method. Experimental results illustrate that the proposed linguistic-base method using ensemble classification attains auspicious performance as compared to state-of-the-art methods.