This Thesis examines the role and limitations of voice biometrics in the contexts of security and for crime reduction. The main thrust of the Thesis is that despite the technical and non-technical hurdles that this research has identified and sought to overcome, voice can be an effective and sustainable biometric if used in the manner proposed here. It is contended that focused and continuous evaluation of the strength of systems within a solid framework is essential to the development and application of voice biometrics and that special attention needs to be paid to human dimensions in system design and prior to deployment. Through an interdisciplinary approach towards the theme reflected in the title several scenarios are presented of the use of voice in security / crime reduction, crime investigation, forensics and surveillance contexts together with issues surrounding their development and implementation. With a greater emphasis on security-oriented voice verification (due to the diversity of the usage scenarios and prospect of use) a new framework is presented for analysis of the reliability and security of voice verification. This research calls not only for a standard evaluation scheme and analytical framework but also takes active steps to evaluate the prototype system within the framework under various conditions. Spoof attacks, noises, coding, distance and channel effects are among the factors that are studied. Moreover, an additional under-researched area, the detection of counterfeit signals, is also explored. While numerous technical and design contributions made in this project are summarised in chapter 2, the research mainly aims to provide solid answers to the high-level strategic questions. The Thesis culminates in a synthesis chapter in which realistic expectations, design requirements and technical limitations of the use of voice for criminological and security applications are outlined and areas for further research are defined.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:528766 |
| Date | January 2010 |
| Creators | Hosseyndoust Foomany, Farbod |
| Contributors | Hirschfield, Alex ; Ingleby, Michael |
| Publisher | University of Huddersfield |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://eprints.hud.ac.uk/id/eprint/9635/ |
Page generated in 0.0013 seconds