• This website provides databases of audio recordings for forensic-voice-comparison research and practice.
  
  
  
• The databases are made available to researchers in order to foster research on forensic voice comparison.
  
  
  
• They are also made available to practitioners in order to assist them to make numeric estimates of likelihood ratios as strength-of-evidence statements in casework, and to test the validity and reliability of forensic-voice-comparison systems under casework conditions.
  
  
  
• We collect multiple non-contemporaneous recordings of each speaker, and in each recording session collect recordings of different speaking styles.
  
  
  
• Details of our data-collection protocol and the rationale behind it are presented in:
  
  
  • Morrison, G. S., Rose, P., & Zhang, C. (2012). Protocol for the collection of databases of recordings for forensic-voice-comparison research and practice. Australian Journal of Forensic Sciences, 44, 155–167.
    doi:10.1080/00450618.2011.630412
    
    
    
• The recordings were made in a high-quality format and have been manually processed to remove extraneous noises and crosstalk.
  
  
• The recordings can be further processed to reflect the conditions of particular cases.

AusEng 500+ (2021 release)

• This database contains 3899 recordings totalling 310 hours of speech from 555 Australian-English speakers.
  
  
  • 324 female speakers
    
    
    • 91 recorded in one recording session
      
      
    • 69 recorder in two separate recording sessions
      
      
    • 159 recorded in three recording sessions
      
      
    • 5 recorded in more than three recording sessions
      
      
  • 231 male speakers
    
    
    • 62 recorded in one recording sessions
      
      
    • 43 recorder in two separate recording sessions
      
      
    • 107 recorded in three recording sessions
      
      
    • 19 recorded in more than three recording sessions
      
      
      
• Each recording session was usually separated by one to two weeks.
  
  
  
• On each occasion each speaker was recorded performing three speaking tasks:
  
  
  • casual telephone conversation (cnv)
    
    
  • information exchange task over the telephone (fax)
    
    
  • pseudo-police-style interview (int)
    
    
    
• The recordings were made using high-quality audio equipment in a quiet environment and have been manually processed to remove extraneous noises and crosstalk.
  
  
  
• Database collection materials
  
  
  • Instructions for speakers, and faxes
    
    
    
• Spreadsheet with details of:
  
  
  • Speaker demographics
    
    
  • Number of recordings per speaker
    
    
  • Length of each recording
    
    
    
• The database was collected by the
  
  Forensic Voice Comparison Laboratory, School of Electrical Engineering & Telecommunications, University of New South Wales (UNSW)
  
  as part of the Australian Research Council funded Linkage Project Making demonstrably valid and reliable forensic voice comparison a practical everyday reality in Australia
  
  in partnership with:
  
  
  • Australian Federal Police
    
    
  • New South Wales Police
    
    
  • Queensland Police
    
    
  • National Institute of Forensic Sciences
    
    
  • Australasian Speech Sciences and Technology Association (ASSTA)
    
    
  • Guardia Civil
    
    
  • Universidad Autónoma de Madrid
    
    
    
• Work leading to the 2021 update was funded by Research England’s Expanding Excellence in England Fund as part of funding for the Aston Institute for Forensic Linguistics 2019–2023.
  
  
  
• The database is available for non-commercial research and forensic casework.
  
  
  • download database (login required)
    
    
  • 
    
    If the automatic login request does not work on your system, use the manual request form.
    
    
    
• Please cite as:
  
  
  • Morrison G.S., Zhang C., Enzinger E., Ochoa F., Bleach D., Johnson M., Folkes B.K., De Souza S., Cummins N., Chow D., Szczekulska A. (2021). Forensic database of voice recordings of 500+ Australian English speakers (AusEng 500+). [Available: http://databases.forensic-voice-comparison.net/]
    
    
    
• History:
  
  
  • 2015-02-24: initial release
    
    
  • 2016-08-12: corrected version – We discovered that the previous version contained several mislabeled folders. These have been manually corrected and checked. If you downloaded before 2016-08-12, we recommend that you discard the earlier version and download the corrected version instead.
    
    
  • 2016-08-19: counting error in spreadsheet corrected
    
    
  • 2016-08-20: folder 2017(3)_cnv moved from female to male folder (this was a missing folder - it was not listed in the spreadsheet)
    
    
  • 2021-07-12 update released 2021-08-02: Additional recordings added (as indicated in spreadsheet). Multiple utterance-length recordings concatenated into one recording per task per session. Resampled from 44.1 kHz to 16 kHz. Distributed in flac format.
    
    
  • 2022-07-25: A few errors in the spreadsheet and in filenames corrected – corrections noted in the spreadsheet.
    
    
    

forensic_eval_01

Multi-laboratory evaluation of forensic voice comparison systems under conditions reflecting those of a real forensic case

• There is increasing pressure on forensic laboratories to validate the performance of forensic analysis systems before they are used to assess strength of evidence for presentation in court (including pressure from the recently released report by the President’s Council of Advisors on Science and Technology, PCAST). Different forensic voice comparison systems may use different approaches, and even among systems using the same general approach there can be substantial differences in operational details. From case to case, the relevant population, speaking styles, and recording conditions can be highly variable, but it is common to have relatively poor recording conditions and mismatches between the known- and questioned-speaker recordings. In order to validate a system intended for use in casework, a forensic laboratory needs to evaluate the degree of validity and reliability of the system under forensically realistic conditions. We have released a set of training and test data representative of the relevant population and reflecting the conditions of an actual forensic voice comparison case, and operational forensic laboratories and research laboratories are invited to use these data to train and test their systems. The details below include the rules for the evaluation, a description of the data, and a description of the evaluation metrics and graphics. The name of the evaluation is: forensic_eval_01
  
  
• Papers reporting on the results of the evaluation of each system were published in a Virtual Special Issue (VSI) of Speech Communication.
  
  
  
• Video introduction to forensic_eval_01 (presented at Odyssey 2016)
  
  
  
• Details including rules for the evaluation (introductory paper for the VSI)
  
  
  • The VSI was completed in June 2019.
    
    
  • The forensic_eval_01 training and test data are now available without the condition that a paper be submitted to the VSI.
    
    
  • The Matlab script to calculate the performance metrics and draw the performance graphics is also now available.
    
    
  • 
    
    Internet Explorer users: To view the automatic login request above please set the security settings to "Medium High" or lower.
    
    If the automatic login request does not work on your system, use the manual request form.
    
    
    
  • download data and Matlab script (login required)
    
    
    

Standard Chinese: 68 female speakers

• This database is no longer distributed.
  
  
  
• This database contains two non-contemporaneous recordings of each of 68 female speakers of Standard Chinese (a.k.a. Mandarin and Putonghua).
  
  
  
• 60 of the speakers are from north eastern China, and 8 are from southern China.
  
  
  
• Each speaker was recorded in three speaking styles:
  
  
  • casual telephone conversation (cnv)
    
    
  • information exchange task over the telephone (fax)
    
    
  • pseudo-police-style interview (int)
    
    
    
• Recordings were made at China Criminal Police University in Shenyang, China, under the supervision of Cuiling Zhang.
  
  
  
• Data collection was financially supported by an International Association for Forensic Phonetics and Acoustics (IAFPA) Research Grant:
  
  
  • Zhang, C., & Morrison, G.S. (2009). Forensic comparison of Chinese female voices. Application for an International Association for Forensic Phonetics and Acoustics Research Grant.
    
  • download pdf
    
    
    
• Database collection materials
  
  
  • Instuctions for speakers, and faxes
    
    
    
• Sound file infomation
  
  
  • Includes suggested split of NE speakers into background, development, and test sets.
    
    
    
• Please cite as:
  
  
  • Zhang, C., & Morrison, G. S. (2011). Forensic database of audio recordings of 68 female speakers of Standard Chinese. [Available: http://databases.forensic-voice-comparison.net/]
    
    
    
• Release dates:
  
  
  • 2011-03-13 original release of high quality audio
    
    
  • 2015-02-16 release of “fax” task in telephone-degraded conditions (see Zhang, Morrison, Enzinger, Ochoa, 2013)
    
    
  • 2025-10-17 database not longer distributed
    
    
    
• Related papers include:
  
  
  • Zhang, C., Morrison, G.S., Enzinger, E., Ochoa, F. (2013). Effects of telephone transmission on the performance of formant-trajectory-based forensic voice comparison – female voices. Speech Communication, 55, 796–813.
    doi:10.1016/j.specom.2013.01.011
    
    
  • Zhang, C., Morrison, G. S., Ochoa, F., Enzinger E. (2013). Reliability of human-supervised formant-trajectory measurement for forensic voice comparison. Journal of the Acoustical Society of America, 133, EL54–EL60.
    doi:10.1121/1.4773223