Evaluation of otoacoustic emissions as a biometric

• Main Authors: Grabham, N.J (Author), Swabey, M.A (Author), Chambers, P. (Author), Lutman, M.E (Author), White, N.M (Author), Chad, J.E (Author), Beeby, S.P (Author)
• Format: Article
• Language: English
• Published: 2013-01-02.
• Subjects: Article
• Online Access: Get fulltext

 This paper presents a comprehensive overview of an investigation into the use of Otoacoustic Emissions (OAE) as an identification biometric. OAE could be important as a biometric identifier in applications where users wear headsets since it is discrete and difficult to spoof. OAE are very low level (~17 dB Sound Pressure Level (SPL)) sounds emitted from the human ear as part of the normal hearing process. They can occur spontaneously or be invoked by a suitable stimulus, these being known as Transient Evoked Otoacoustic Emissions (TEOAE) and Distortion Product Otoacoustic Emission (DPOAE). An initial visual comparison shows that otoacoustic emissions are clearly distinctive and are stable over a six month period. A biometric analysis based on the Euclidean distance measurement of TEOAE recordings in the temporal domain was performed on pre-recorded datasets captured for medical purposes and data were collected specifically for this study. For a database of 23 subjects, the predicted Equal Error Rate (EER) was 1.24% for a 90% confidence interval. DPOAEs also demonstrated biometric potential but the level of discrimination is inferior to TEOAE. The combination of DPOAE and TEOAE into a multimodal analysis was demonstrated to be feasible although the potential improvement in performance is yet to be quantified. Finally the use of Maximum Length Sequencing (MLS) was investigated to reduce capture time without decreasing performance. This demonstrated a reduction in capture time for a TEOAE from 1 minute to 5 seconds with a visual analysis of a 4th order MLS showing good stability and reproducibility. OAEs can potentially be used as a biometric and benefit from their small template size (512 data points in our TEOAE biometric) and simple analysis. The level of background noise is the most significant practical factor that affects biometric performance.