Application－Forensic Speaker Recognition  

Forensic speaker recognition (FSR) is the forensic application of speaker recognition technology. Actually, FSR is the decision-making process during which one or more samples of an unknown voice will be compared with one or more samples of a known voice and then a decision whether they are of the same origin or not has to be made.

So far, forensic phoneticians and speech signal processing engineers are the main researchers in this field, whose achievements have contributed to the current diverse FSR methods in practice.

However, there has been being a gap between ‘unrealistic test material’ in research and real voice materials involved in legal practice; real-world conditions exercise ‘drastic effect’ on FSR (Rose, 1996; 2002:92); and evaluation and validation of different methods are absent ( Cambier-Langeveld, 2007). It seems apparent that the current FSR research and development have fallen behind the demands of legal practice.

Guan Xin (2014) proposes a new perspective to conduct FSR research, which attempts to observe the audio material as a whole composed of speech sound and speech content. Such a proposal is theoretically reasonable, while its practical feasibility needs proving. Based on the proposal, studies are underway to test the operability of the proposed method.

 

References:

1. Rose, P. 1996. Speaker Verification under Realistic Forensic Conditions[A]. Proceedings of the Sixth Australian International Conference on Speech Science and Technology, Australian Speech Science and Technology Association, Canberra, 109-114.

2. Rose, P. 2002. Forensic Speaker Identification [M]. London & New York: Taylor & Francis.

3.Cambier-Langeveld, T. 2007. Current methods in forensic speaker identification: Results of a collaborative exercise [J]. International Journal of Speech Language and the Law 14(2): 223-243.

4. Guan, Xin. 2014 (in press). Study on FSR cross-validation method [J].