On Applicability of Tunable Filter Bank Based Feature for Ear Biometrics: A Study from Constrained to Unconstrained

Area of Science:

• Biometrics
• Image Processing
• Pattern Recognition

Background:

• Person verification is crucial for security.
• Ear biometrics offers a unique and stable identification method.
• Traditional feature extraction methods have limitations.

Purpose of the Study:

• To present a framework for person verification using ear biometrics.
• To evaluate the effectiveness of a tunable filter bank as a local feature extractor for ear images.
• To assess the performance of the proposed method on diverse ear image databases.

Main Methods:

• A tunable filter bank, based on a 14th order half-band polynomial, was employed for feature extraction.
• Distinct features were extracted from ear images while preserving frequency selectivity.
• Recognition tests were conducted on constrained (AMI, WPUT, IITD) and unconstrained (UERC) ear image databases.
• Experiments involved analyzing four and six subdivisions of ear images.

Main Results:

• The tunable filter bank demonstrated moderate success in distinguishing ear features.
• Recognition accuracies achieved were 70.58% (AMI), 67.01% (WPUT), 81.98% (IITD), and 57.75% (UERC).
• Canberra Distance was used as the underlying measure of separation.

Conclusions:

• The tunable filter bank is a viable candidate for person recognition using ear biometrics.
• The method shows potential for distinguishing ear features effectively.
• Further research can explore optimizations for improved accuracy across various conditions.