Congruence plots and their applications in forensic voice comparison

Kelly F, Coy T, Alexander A, Jessen M. Congruence plots and their applications in forensic voice comparison. In: AES 2024 International Conference on Audio Forensics. Audio Engineering Society; 2024. Paper 5. Available from: https://aes.org/publications/elibrary-page/?id=22630

@inproceedings{Kelly2024_22630,
  author = {Kelly, Finnian and Coy, Thomas and Alexander, Anil and Jessen, Michael},
  title = {{Congruence plots and their applications in forensic voice comparison}},
  booktitle = {AES 2024 International Conference on Audio Forensics},
  note = {Paper 5},
  year = {2024},
  month = jun,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=22630}
}

TY  - CPAPER
TI  - Congruence plots and their applications in forensic voice comparison
AU  - Kelly, Finnian
AU  - Coy, Thomas
AU  - Alexander, Anil
AU  - Jessen, Michael
T2  - AES 2024 International Conference on Audio Forensics
M1  - Paper 5
PY  - 2024
DA  - 2024/06/17
UR  - https://aes.org/publications/elibrary-page/?id=22630
PB  - Audio Engineering Society
LA  - en
AB  - In forensic voice comparison, it is common for a practitioner to compare the same set of recordings using multiple speaker recognition systems or methods. The usual approach for comparing the results obtained from different systems or methods is to use performance metrics such as the Equal Error Rate (EER) or log-likelihood ratio cost (Cllr), and graphics like the Detection Error Tradeoff (DET) curve and Tippett Plot. While these metrics and graphics are very useful indicators of overall system discrimination and calibration, they do not indicate the level of agreement, or congruence, between the systems at the level of individual comparisons; for example, they do not inform as to whether System A and System B provide the same categorical output for an individual comparison. The congruence plot provides a visual representation of system agreement on a comparison-by-comparison basis by plotting the scores obtained for the same set of comparisons from two different systems against one another. This paper will introduce the congruence plot and its features within the Bio-Metrics software tool. The applications of congruence plots for voice comparison research and casework will then be illustrated through three forensically-relevant case studies.
ER  -