ELEC9781
Forensic Voice Comparison and the Evaluation of Evidence
This is a 6 credit UNSW graduate course.
The first part of this course is open to auditors at no charge. Please contact Dr Morrison if you are interested in auditing the first part of the course.
Times
- First part:
- Monday 1 August – Friday 5 August 2011
- Monday 8 August – Friday 12 August 2011
- Second part:
- Thursday 25 August – Thursday 20 October 2011
- every Thursday, 1:00pm–3:00pm
Instructors
- Dr Geoffrey Stewart Morrison
- Dr Julien Epps
Description
- This course introduces the theory and practice of forensic voice comparison conducted within the new paradigm for the evaluation of forensic evidence.
- The first part of the course (taught primarily by Morrison) focusses on the new paradigm for the evaluation of forensic evidence and its application to forensic voice comparison. The new paradigm is characterised by the quantitative implementation of the likelihood-ratio framework using databases and objective measurements, and empirical testing of the validity and reliability of forensic-comparison systems under conditions reflecting those of the case at trial. The likelihood-ratio framework is the logically correct framework for the evaluation of evidence and was adopted as standard for forensic DNA-profile comparison in the mid 1990s. Forensic voice comparison plays a leading rôle in the adoption of the new paradigm among other branches of forensic science.
- The first part of the course should be of interest and accessible to a broad audience, with the concepts introduced applicable across all branches of forensic-comparison science and pertinent to the admissibility of forensic evidence in general. Interested lawyers, judges, police officers, and forensic scientists working in other branches of forensic science are welcome to audit this part of the course, which is taught in intensive mode.
- The second part of the course (taught primarily by Epps) focusses on techniques from automatic speaker recognition and their application to forensic voice comparison. A number of automatic methods relevant to forensic voice comparison will be covered, including feature extraction for automatic speaker characterization (acoustic, dynamic, prosodic, linguistic), feature normalization, speaker modelling using Gaussian mixture models, speaker adaptation and GMM supervectors, channel compensation methods, and automatic segment selection.
- The course will provide students with a solid grounding in the new paradigm for the evaluation of forensic evidence and advance students’ technical skills and knowledge towards the state of the art in acoustic-phonetic and automatic approaches to forensic voice comparison.
Additional information
ELEC9782
Advanced Forensic Voice Comparison and the Evaluation of Evidence
This is a 6 credit UNSW graduate course officially taught Semester 2, 2012.
Instructor
Description
- ELEC9781 “Forensic Voice Comparison and the Evaluation of Evidence” introduced the theory and practice of forensic voice comparison conducted within the new paradigm for the evaluation of forensic evidence. The first part of the course focussed on the new paradigm for the evaluation of forensic evidence and its application to forensic voice comparison. The second part of the course focussed on techniques from automatic speaker recognition and their application to forensic voice comparison.
- ELEC9782 “Advanced Forensic Voice Comparison and the Evaluation of Evidence” explores these topics in greater depth and breadth via extensive reading of primary literature and seminar discussions of selected important works. Whereas the readings in the first part of ELEC9781 were primarily authored by Morrison, the readings in ELEC9782 are primarily by other authors.
- The course will provide students with an advanced understanding of the new paradigm for the evaluation of forensic evidence, its application to forensic voice comparison, and related topics in the evaluation of forensic evidence.
Additional information