Theoretical considerations in creating comprehensive computerized speaker profiling

Speaker profiling is the art and science of making inferences about a subject or suspect based on "what" he or she says and "how" it is spoken. It is part of the larger discipline of "criminal-behavioral profiling" used by police, military, homeland security, and intelligence organizations with widespread application. This paper examines the evolving art and science of speaker profiling and provides the theoretical basis for computerized analysis of speech patterns. We propose that supporting science and current technology exists to develop automatic speaker profiling which can provide rapid comprehensive supplements to criminal-behavioral profiles for use by the military, intelligence agencies, homeland security, police departments, and at ports-of-entry. Theoretically, there are five areas where automatic computerized speaker profiling can provide valuable criminal-behavioral information: accent-dialect analysis, intoxication investigation, lie detection and stress analysis, and speaker identification. When considering the above aspects of speaker profiles, a clear distinction is made between information based on science and the conjecture that can be drawn from them. In this paper, the theoretical bases for computerized speaker profiling is provided.