Ear Recognition Algorithm Research Based On DWT, PCA And LDA

From the nineties to now, with the rapid development of the computing networktechnology, people’s daily lives toward a direction of comprehensive electronic and automatic.biometrics identification technology as an important branch in pattern recognition field, haveplayed an important role in people’s lives, such as scientific research, medical diagnostics,aerospace, agriculture, transportation, military, media and so on. Each biometricsidentification technology has its inherent disadvantages and limitations, which resulting in notachieving the desired effect. Human ear as an aspect of biometric identification has highresearch value.At present, ear recognition research in China is still in laboratory stage. The human earrecognition system including four aspects, such as ear image acquisition, human ear imagepreprocessing, ear feature extraction and the characteristics of human ear recognition, thefocus of this paper is to study ear feature extraction and feature recognition.This paper first introduces the significance of the human ear recognition research, andthe development status at home and abroad, and the advantages of human ear’s identityaccording to the ear structure feature, then briefly introduces the human ear recognitionsystem, human ear image preprocessing and the human ear detection, and then lists severalperformance evaluation of ear recognition, At last this paper introduces principal componentanalysis(PCA), linear discriminant analysis(LDA)and knowledge of discrete wavelettransform(DWT) and so on. Finally, propose ear recognition algorithm based on DWT, PCAand LDA and made simulation experiments, and compare it with the classical ear recognitionalgorithms based on PCA and LDA.Human ear recognition algorithms based on DWT, PCA and LDA firstly uses wavelettransform algorithm to extract the human ear image low-frequency information, and then usesprincipal component analysis to extract the main component of the human ear image. Finally,uses linear discriminant analysis to obtain optimal ample space mapping and classifies themwith nearest neighbor rule. Experiments show that this method can overcome the influence ofillumination to the experiment and the recognition rate is up to98.33%under the test ofUSTB ear image library, higher than the recognition rate of96.67%which based on classicalalgorithms based on PCA and LDA, but this method is bad when the angle of the picturechanged, we need make further study to propose new solutions.