Human perception based criminal identification through human robot interaction

Abstract

In India like many other countries, the crimes especially against women are rising. One of the many reasons perhaps is the low conviction rate. The initial stage of criminal investigation starts with the exploration of evidences and eyewitnesses. An eyewitness can act as a guide to trace out the suspect. Her/his perception about the suspect can be useful to identify the criminal. Based on the descriptions of the eyewitnesses normally a manual sketch is prepared and released in the Newspapers which most of the times is vague and ambiguous and hence ineffective. Therefore, a robotic framework has been proposed in this paper to help Police to identify criminals using the imprecise knowledge about the subject. The designed robotic system interacts with the eyewitness by asking several questions about the suspect such as age, height, her/his facial shape and size etc., and then making a guess about her/his face. A human face can be considered as the combination of various macro facial features such as eye. Eyebrows, face shape, lip shape, nose type etc. Although these features vary from person to person but their combination makes a human face unique. An experimental study on 113 Indian Bollywood celebrities, 37 Indian cricketers and 40 persons from Robotics and Artificial Intelligence Laboratory of Indian Institute of Technology Allahabad, India has been performed. The system is able to identify the criminal (here pseudo criminals) if it exists in the database. This system can be useful at the initial stage of investigation where we don't have any knowledge about the criminal.