Ear  Recognition  System   

Biometrics identification methods have been proved to be very efficient, more natural and easy for users than traditional methods of human identification. Among various physiological biometrics, ear biometrics has been found to be interesting and promising in the recent past. The ear of human goes through minute changes as course of age. Its also has no changes as expression change. So ear biometrics emerges as an efficient biometric method for human identification and could be used like other biometrics. Database for a ear based biometric system is prepared by processing images of the ears of authorized personnel, followed by extraction of characteristic features for each image. Personnel wishing to enter would have their ears focused to the camera lens at the entrance and the image would be processed and compared for a match against the database. The stored features would have to be sufficiently distinct so as to be able to distinguish one ear from all the others, and sufficiently robust so that the same features would be produced every time the ear image is taken. These are conflicting requirements and present a challenge to the system designer. Ears have certain advantages over the more established biometric traits such as face, fingerprint etc; they have a rich and stable structure that is preserved well even at the old age. The ear does not suffer from changes in facial expression and is firmly fixed in the middle of the side of the head so that the immediate background is predictable. Collection does not have any associated problems, as may be the case with direct contact fingerprint scanning as well as may happen with iris and retina measurements. The size of the ear is large compared with the iris, retina, and fingerprint and therefore is more easily captured. 

Like other biometric systems, an ear recognition system can have two main modules: database creation module and verification module. The database of authorized person is prepared by processing images of the ear. It extracts features for each image and stores in the database. At the time of verification features from the query image of a person are extracted and finally are compared with the corresponding features of the person. Thus the features have to be sufficiently distinct so as to be able to distinguish one ear from all the others, and sufficiently robust so that the same features would be produced every time the ear image is taken. These requirements present a challenge to the system designer.

Our Ear Recognition Approach

Our team implemented an efficient ear recognition system for human identification based on unique pattern of the ear. The proposed method extracts wavelet coefficient matrix to interpret the pattern of the ear as unique feature. In this approach the Haar wavelet transform decomposes the ear images and computes co-efficient matrix of wavelets of ear which are unique. As the Haar wavelet is very sensitive on orientation change of the images, the each image is normalized in fixed orientation as well as intensities are adjusted to make images illumination invariant. The approach has been tested on IITK database which has 750 ear images. It has shown that accuracy of 96.8% with 2.23% FAR and 4.00% FRR.

Automatic Ear Detection

To automate the ear based recognition process, ear need to be detected automatically. We are working on the development of a robust technique which can detect ear correctly and efficiently from the side face image. Automatic ear detection is used as a preprocessing step which takes raw input image of side face and detects and crops the ear. Cropped ear  is passed to main recognition module where features are extracted and matching is performed.

Papers related to Automatic Ear Detection can be found here.