Personal Identification Based on the Individual Sonographic Properties of the Auricle Using Cepstral Analysis and Bayes Formula

• Published in: Cybernetics and systems analysis Vol. 57; no. 3; pp. 455 - 462
• Main Authors: Sulavko, A. E., Lozhnikov, P. S., Kuprik, I. A., Samotuga, A. E.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2021; Springer Nature B.V
• Subjects: Artificial Intelligence; Bayesian analysis; Biometric identification; Biometrics; Cepstral analysis; Classifiers; Control; Ear; Machine learning; Mathematics; Mathematics and Statistics; Parameters; Processor Architectures; Recognition; Software Engineering/Programming and Operating Systems; Systems Theory; Window functions; acoustic signal; windowed Fourier transform; machine learning; cepstrograms; Bayes theorem; pattern recognition
• ISSN: 1060-0396; 1573-8337
• DOI: 10.1007/s10559-021-00370-w

A method of personality recognition by echographic parameters of the human ear is developed based on the naive Bayes classifier in the two following modes: the biometric identification (EER = 0.0053) and the biometric authentication (FRR = 0.0002 at FAR ≤ 0.0001), respectively. A device is developed for recording the biometric characteristics of the external ear, and a set of echographic data is collected from the external ears of 75 subjects. The spectral and cepstral characteristics of the signals reflected from the ear canal are used as biometric parameters. Several window functions for constructing spectra and cepstrograms are considered. It is established that more than 90% of “cepstral” features have a weak correlation, which allows us to use the naive Bayesian classifier and to obtain highly accurate results of user recognition at the same time. An advantage of the Bayesian classification is the possibility of the robust fast learning of the identification system.