Biometric identification based on Transient Evoked Otoacoustic Emission

Biometrics provides a reliable and efficient solution to identity management in many aspects of daily lives, such as application login, access control and transaction security. This paper presents a novel approach to individual identification based on a new biometric modality Transient Evoked Otoacoustic Emission (TEOAE), which is a low level acoustic signal generated by human cochlea and detected in the outer ear canal. We resort to wavelet analysis to derive the time-frequency representation of such non-stationary signal and machine learning techniques: linear discriminant analysis and softmax regression to accomplish pattern recognition. We also introduce a complete framework of the biometric system considering practical application. Experiments on a TEOAE dataset of biometric setting show the merits of the proposed method. With fusion of information from both ears an average identification rate 98.72% is achieved.