Abstract—In this paper a novel algorithm the accuracy of finger vein recognition. Principal Component Analysis (PCA), Kernel Analysis (KPCA), and Kernel Entropy Compo in this algorithm are validated and compared w to determine which one is the most appropriate vein recognition. Keywords—Biometrics, finger vein r Component Analysis (PCA), Kernel Principa (KPCA), Kernel Entropy Component Analysis I. INTRODUCTIO HE importance of verification and gained a lot of attention as the numbe not willing to be identified is getting l passing the time[1]; illegal immigrants, c the run can be two clear examples. Finger recently proposed method in which fing analyzed. Based on the physical charac patterns, finger vein is unique and individ used in finger vein recognition is ‘ recognition algorithms[2] have already b finger vein database that were bo unsuccessful. Principal Component Ana commonly used method for pattern e recognition. Based on previous researc Component Analysis (KPCA)[4], and Component Analysis (KECA)[5] were p the performance of PCA. KPCA is an exte is a linear method which extracts the feat dimension. In case of KPCA, however KPCA the input data is first nonlinearly feature space called F, and then a PCA w the mapped dataset. This nonlinearly m data is done by a function called Φ Component Analysis is exactly the same Component Analysis except for one point to choosing the eigenvectors to project o KPCA (in which the top eigenvecto KECA[6] the chosen eigenvectors have entropy estimate of the input data. Authors are with Intelligent Biometric Group, Electronic Engineering,Universiti Sains Malaysia, U 14300 Nibong Tebal, Pulau Pinang, sepehr125@yahoo.com, Ali_khalili1363@yahoo.co cherish.gjp@gmail.com, m saba_nazariorif@yahoo.com, shad mohammadali_bagheri@yahoo.com. T m is proposed to merit The performances of l Principal Component onent Analysis (KECA) with each other in order e one in terms of finger recognition, Principal al Component Analysis (KPCA). ON d identification has er of people who are larger and larger by criminals who are on vein recognition is a ger vein patterns are cteristics of the vein dual. As the database ‘image’, some face been experienced on oth successful and alysis (PCA)[3] is a extraction and face ch, Kernel Principal d Kernel Entropy proposed to enhance ension of PCA. PCA tures and reduces the r, it is different. In y mapped to another will be performed on mapping of the input Φ. Kernel Entropy e as Kernel Principal t that when it comes onto the data, unlike ors are chosen), in to contribute to the School of Electrical and USM Engineering Campus, Malaysia. E-mails : om, pashna@gmail.com, mehranpower@yahoo.com, di.mahmoodi@yahoo.com, In Section 2, Image acqui finger vein recognition algor Kernel Principal Component Section 5, Kernel Entropy C explained. In section 6, exp database are given. Finally, se II. IMA Deoxygenated haemoglobi rays’ .based on this pro coefficient (AC) of the vein i when you are capturing th following four low-cost pr capture the finger vein data vein images, a microcompu control the LED array, a su control circuit (in this pape emitting diodes (IR LED) w of the constructed capturing to capture finger vein image at the bottom of the design) cam blocks the infrared rays blocking filter. Hnce, to make rays, the negative film is em filter in order to capture the negative film can operat wavelength of 850nm as an the used devices to capture t between them. Fig. 1 used devices to capture fin III. FINGER VEIN This section explains the algorithm in this work. The f shown in Figure 2. Based on five steps: first step is extrac The second step is enhancin limited adaptive histogram eq of images is implemented in s the PCA, KPCA [7] and reducing the dimension of comparing test and train data isition is explained. In Section 3, rithm is introduced. In Section 4, Analysis is explained briefly. In Component Analysis (KECA) is perimental results on finger vein ection 7 concludes the paper. AGE ACQUISITION in in the vein can absorb the light oven scientific; the absorption is higher than other parts of finger he image of finger vein. The rototype devices were used to abase: a computer to process the uter unit (MCU) to adjust and uitable infrared LED and related er Osram SFH485 infrared light ith wavelength 880nm at the top g model has been used), a camera es (Logitech V-UAV35 web-cam ). The Logitech V-UAV35 web- (IR), because it consists of an IR e this camera sensitive to infrared mployed instead of the blocking e transmitted infrared rays. This te to transmit 90% of radiation IR pass filter. Figure 1 indicates the database and the connections nger vein images and the connections RECOGNITION ALGORITHM e used finger vein recognition flow diagram of the finger vein is n the diagram, this algorithm has cting the region of interest (ROI). ng the images by using contrast qualization. Feature normalization step three. Step four is employing KECA to extract features, and f the images. The last step is using Euclidian distance. Sepehr Damavandinejadmonfared, Ali Khalili Mobarakeh, MMoohhsseenn PPaasshhnnaa,, JJiiaannggppiinngg Gou Sayedmehran Mirsafaie Rizi, Saba Nazari, Shadi Mahmoodi Khaniabadi, Mo ohhaammad Ali Bagheri Finger Vein Recognition ussiin ngg PPCCAA-based Methods World Academy of Science, Engineering and Technology 66 2012 1079