Abstract
In certain cases, the only evidence to identify terrorists, who are seen in digital images or videos is their hands’ shapes, particularly, the victory sign as performed by many of them when they intentionally hide their faces, and/or distort their voices. This paper proposes new methods to identify those persons for the first time from their victory sign. These methods are based on features extracted from the fingers areas using shape moments in addition to other features related to fingers contours. To evaluate the proposed methods and to show the feasibility of this study we have created a victory sign database for 400 volunteers using a mobile phone camera. The experimental results using different classifiers show encouraging identification results; as the best precision/recall were achieved by merging normalized features from both methods using linear discriminate analysis classifier with 96.6% precision and 96.3 recall. Such a high performance achieved by the proposed methods shows their great potential to be applied for terrorists’ identification from their victory sign.