Iris Recognition Based on 2D Wavelet and AdaBoost Neural Network

Biometrics refers to automatic identity authentication of a person on the basis of one's unique physiological or behavioral characteristics. To date, many biometric features have been applied to individual authentication. The iris, a kind of physiological feature with genetic independence, contains an extremely information-rich physical structure and unique texture pattern, and thus is highly complex enough to be used as a biometric signature. Statistical analysis reveals that irises have an exceptionally high degree-of-freedom up to 266 (fingerprints show about 78) [1], and thus are the most mathematically unique feature of the human body, more unique than fingerprints. Hence, the human iris promises to deliver a high level of uniqueness for authentication applications that other biometrics cannot match.

The outline of this chapter is as follows. The method that uses a 2-D wavelet transform to obtain a low-resolution image and a Canny transform to localize pupil position is presented in Sect. 8.2. By the center of the pupil and its radius, we can acquire the iris circular ring. Section 8.3 adopts the Canny transform to extract iris texture in the iris circular ring as feature vectors and vertical projection to obtain a 1-D energy signal. The wavelet probabilistic neural network is a very simple classifier model that has been used as an iris biometric classifier and is introduced in Sect. 8.4. Two different extension techniques are used: wavelet packets versus Gabor wavelets. The wavelet probabilistic neural network can compress the input data into a small number of coefficients and the proposed wavelet probabilistic neural network is trained by the AdaBoost algorithm. The experimental results acquired by the method are presented in this section. Finally, some conclusions and proposed future work can be found in Sect. 8.8.