ABSTRACT
A new hierarchical scheme for criminal investigation, based on multi-sensory data including voice, fingerprint, facial image and incidental description, has been outlined in this chapter. The proposed system first attempts to classify a given fingerprint of a suspect into any of the six typical classes and then determines the fingerprint that matches best under the class. In case the fingerprint matching fails, an automated scheme for matching facial images is called for. On successful match, the suspect could be identified. In case the matching results in poor performance, the speaker identification scheme that works with phoneme analysis is utilized. When no conclusion about the suspect could be detected by voice, incidental description is used as the resource for criminal investigation. The incidental description based model is realized on a Petri net that can handle imprecision and inconsistency of data and uncertainty of knowledge about the criminology problem.
