Identifying Criminal Suspects by Human Gestures Using Deep Learning

When it comes to investigations, suspects are well aware of the questions asked to them and know every possible way to answer them without getting caught But in the aspect of behaviour and attitude, each person is unique and it can be identified through body language, and hand gestures. Gestures are used in criminology departments to identify criminals based on body language, eye contact and hand gestures. Since the above process is manually done by the investigators, they begin to lose the ability to identify the suspects by investigating throughout the day. To assist police personnel during the investigation, a system is proposed which automatically identifies the suspects with deep learning techniques based on gestures. Gestures are found to be appealing as they are a part of how humans communicate. The proposed system follows the standards of Federal Bureau of Investigation (FBI) former agent which helps to read body languages based on different types of positive and negative hand gestures. Convolutional Neural Networks (CNNs) utilized in the proposed system is capable of learning and classifying multifarious hand gestures at lower error levels. With the dataset of negative and positive gestures, the networks are well trained and tested, and the model gives an accuracy level of 98.96%.