Biometric Identification using Human Ear
Aishna Sharma1, Natasha Lalwani2, Mani Roja M. Edinburgh3
1Aishna Sharma, Department of Electronics and Telecommunication Engineering, Thadomal Shahani Engineering College, Mumbai, India.
2Natasha Lalwani, Department of Electronics and Telecommunication Engineering, Thadomal Shahani Engineering College, Mumbai, India.
3Dr. Mani Roja M. Edinburgh, Department of Electronics and Telecommunication Engineering, Thadomal Shahani Engineering College, Mumbai, India.
Manuscript received on September 23, 2019. | Revised Manuscript received on October 15, 2019. | Manuscript published on October 30, 2019. | PP: 4893-4898 | Volume-9 Issue-1, October 2019 | Retrieval Number: A2027109119/2019©BEIESP | DOI: 10.35940/ijeat.A2027.109119
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Biometrics refers to the metrics of the human characteristics which has gained much popularity in recent times. It is a form of identification and access control. Widely used forms of biometrics are facial recognition, finger print recognition, iris recognition, etc. but the drawback is that most of these features change over time. The human ear is a cogent source of data to classify biometrically since its attributes do not change substantially as time progresses. This paper explores the field of ear biometric wherein the database images are re-sized to 128 x 256 pixels and then converted to grayscale image. Various transforms viz. Discrete Cosine Transform, Discrete Fourier Transform, Discrete Wavelet Transform are then applied to extract the features. The coefficients of the test image are compared with the coefficients of the registered database image. On comparison, Euclidean distance classifier is used to recognize the test image from the database. The database used consists of 25 subjects with 6 images per person out of which the initial 4 images are used to train the model, and the remaining 2 for testing. The outputs of various transforms were compared and the best accuracy obtained is 86% using Discrete Wavelet Transform.
Keywords: Discrete Cosine Transform, Discrete Fourier Transform, Discrete Wavelet Transform, Ear Recognition, Person Identification