Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.635810
Title: Text-independent speaker verification with limited resources
Author: Auckenthaler, R. W.
Awarding Body: University of Wales Swansea
Current Institution: Swansea University
Date of Award: 2001
Availability of Full Text:
Access from EThOS:
Abstract:
Access to data or secure areas today requires verification of the person's identity. This is commonly performed by entering the PIN, a publicly unknown personal identification number. A person must remember this PIN and access is denied when a wrong number is entered or the number is forgotten. The PIN can also accidentally become known to the public domain. Then every person is able to access the private information. Biological characteristics of a person are used in future applications for the purpose of verification. These "biometrics" increase the convenience of the system and can lead to a higher security against fraud. The latter is due to a harder task of acquiring and reproducing the biometrics of a person, One type of biometrics is the voice, the corresponding recognition task is called speaker verification. Speaker verification can be useful for future applications in wireless devices to secure financial transactions via voice. The use of wireless devices constrains the available resources for speaker verification. Systems in research often use large amounts of memory and processing. A suitable implementation of speaker verification in wireless devices must therefore be optimised with regards to memory and processing. Different approaches to reduce memory and processing requirements are discussed in this thesis. Novel techniques are revealed which lead to significant reductions without degradation in verification performance. With these methods the memory requirement can be reduced by a factor of four. The processing requirement is reduced by a factor of up to forty. Implementing these reduction methods allows a very resource efficient implementation of speaker verification in wireless devices.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (M.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.635810  DOI: Not available
Share: