Voice biometrics could crack down on crime

Dr Aladdin Ariyaeeinia at the University’s School of Electronic Communication & Electrical Engineering has been conducting research into voice biometrics (speaker recognition) for over 10 years. The process has various potential applications such as verifying individuals’ identities when they try to access cash machines or try to bank or shop online.

“The technique we have developed works well in different environments because it encompasses methods to reduce the effects of background noise,” said Dr Ariyaeeinia.

“The process of verifying a claimant will result in a match score. This represents the similarity of the user’s voice to the reference material for the identity being claimed. The University’s method operates in such a way that when the claimant is an impostor, his/her verification score will be suppressed in relation to those for genuine claimants.”

The research in the above area has been the basis for the User Voice Identification (URVIN) project at the University. This is funded by DTI under the New Wave Technologies and Markets programme, and is conducted in collaboration with industry.

Another area of investigation in Dr Ariyaeeinia’s group is the use of voice biometrics for the indexation of conversational audio. This project (part funded by EPSRC) has already led to the development of effective methods for audio data navigation and retrieval. The technique has valuable applications in managing audio-visual recordings, and in processing audio surveillance material for the purpose of combating crime.

Dr Ariyaeeinia commented: “Voice biometrics has the advantage of being useful in areas where you cannot use other biometrics. For instance, the increase in the use of online and telephone banking is expected to lead to the deployment of voice biometrics as an additional means of security.”

Dr Ariyaeeinia’s team is also looking at the fusion of different types of biometrics. The primary purpose of this work on multimodal biometrics is to achieve a level of reliability beyond that obtainable with individual types of biometrics.