The palm vein authentication technology is extremely safe, accurate and reliable as it uses the vascular patterns contained within the body to confirm personal identification. The pattern of veins in the palm is complex and unique to each individual. Its non-contact function gives it a healthful advantage over other biometric technologies. This paper presents an algebraic method for personal authentication and identification using internal contactless palm vein images. We use MATLAB image processing toolbox to enhance the palm vein images and employ coset decomposition concept to store and identify the encoded palm vein feature vectors. Experimental evidence shows the validation and influence of the proposed approach.
Biometrics can be defined as an automated measurement of physiological or behavioral characteristics that are used to authenticate, determine or confirm identity of individuals [
Several important issues have to be taken into consideration to design an effective biometric system [
One of the upcoming highly secure, accurate and precisebiometric technologies is the personal identification using palm veins [
Similar to the problem of obtaining and reviewing a message through a noisy channel, when a user wants to access a system, the access device should permit access when palm vein information does not vary by more than a definite number of binary digits. Also, for security reason, it is eligible that palm vein information is saved in a database (on a token or on a smart card) in encrypted arrangement, rather than in normal digital image. Therefore, the techniques which employ coding theory can be proposed to solve the secure palm vein storage problem. In the literature there are several schemes which use ideas from coding theory for the biometrics storage problem; see for instance [
At the present time many types of biometric systems are in use; the most widespread ones are fingerprints, face recognition, iris recognition, hand and finger geometry, voice print, signature identification, gait, DNA and dorsal hand vein. In the rest this section we will present some of the most commonly used biometric technologies.
The fingerprint recognition devices for laptop and cell phone access are now widely available from many different vendors at a low cost. With fingerprint technology [
The identification of a person by their facial image [
Iris (of the eye) recognition method [
Hand and finger geometry of personal verification [
Voice recognition [
Signature identification is the analyses of the way a user signs his/her name [
A palm has wide area and more complicated vascular pattern and contains some differentiating features which can be used for personal identification. This technology relies on comparing the palm vein patterns present in the ventral side and the dorsal side of both hands and fingers. The idea of using palm vein can be summarized as follows: Hemoglobin in the blood is oxygenated in the lungs and carries oxygen to the tissues of the body through the arteries. After it releases its oxygen to the tissues, the deoxidized hemoglobin returns to the heart through the veins. These two types of hemoglobin have different rates of absorbency. Deoxidized hemoglobin absorbs light at a wavelength of rang 750 nm to 940 nm in the near-infrared region. It has been recently attracted the attention of researchers. It is more preferable than the other types of biometrics as it is impossible to steal or counterfeit the patterns and the pattern of the vessels of the hand is fixed and unique with repeatable biometric features. The system consists of a small palm vein scanner (camera or sensor) that is easy and natural to use, fast and highly accurate. Also there is no need of sensor cleaning. The recent researches have been obtained a certain figure of palm vein recognition rates, see for example [
Our palm vein authentication system is factually composed of five consolidated components. The first component is a near-infrared camera which is used to capture images and to convert them into digital formats. The second component is a signal processing procedure which develops the palm vein template and performs quality control activities. The third component is a data storage which is responsible for keeping information of registered palm vein templates as encoded feature vectors at which any new captured palm vein can be compared to during matching phase. The forth component is a matching algebraic procedure which compares the new palm vein data to one or more palm vein data kept in database storage. The latter component is a decision process which uses the results obtained from the matching component to produce a system-level decision.
There are various civilian areas in which infrared camera can be exploited such as security systems and warning for a threats’ presence. There are two main types of infrared cameras, see [
It is known that each object emits a certain amount of black bodyradiation as a function of its temperature. The higher an object’s temperature, the more infrared radiation is emitted as a black body radiation. A near-infrared camera can reveal this radiation in a way like to the way a customary camera detects the visible light. It can work even in complete darkness because the surrounding light level does not concern. This makes it valuable for rescue operations in undergrounds or smoke-filled buildings. Sensor emanates infrared ray to the persons’ palm which should be placed few centimeters from the sensor. The idea of image capturing is simple and can be summarized as follow: Oxygen reduced blood within the veins absorbs the near-infrared light and then the infrared camera takes a raw image of the palm vein pattern. Then we can precisely identify the structure of the pattern of veins on the palm of the human hand. By using signal processing, the system changes the raw images into palm vein patterns.
The image is captured, by using a specific near-infrared camera, as an input image and various mathematical operations are applied to this image in order to extract all efficient features. The objective of pre-processing is to improve the image information by suppressing unsought distortions and enhancing some image options. This is extremely important for further processing [
In general, there are different image segmentation methods such as region growing method, relaxation method, edge detection method, division and combination method, and threshold method. The dynamic threshold based segmentation process is carried out which subdivides each image into its constituent regions. The palm vein inner layer are extracted according to the specified (or selected) threshold and the segmented palm vein patterns are obtained.
After the pre-processing and obtaining enhanced and segmented palm vein patterns, we use the local line binary pattern (LLBP) method [
The palm vein images which have resolutions 480 × 640 pixels are acquired for various users. These images were used to extract palm vein templates as described in the previous section, then encoded using a suitable code (see below) and stored as code words in the system database.
In matching process if the population is considerable, it might take an extended time to explore through the system database. Thus, to improve the performance of matching system, a common strategy can be used. We may divide the database into various bins at which each bin contains only the palm veins of the same class. In each class we put a representative of the most characteristic information in the palm vein images set in order to reduce the matching time and the size of database. When a palm vein is asked to be identified, it is only compared with these in the bin of the same class. However, dividing the system database into many bins does not always improve the matching system performance. Many realistic systems make use of other features to further divide the database into more bins. Other systems attach palm veins with a number of attributes and classify them according to tags.
Then the system is ready to match palm vein images using the coset decomposition algorithm as a modified hamming distance measure.
The consequence of the matching process between two palm vein feature vectors, having the appearance of being different, could be a decision of either match or no match. Since, the result of the matching is required to be distinctly decision, a reasonable threshold or an acceptable difference level is demanded. Therefore, an algorithmic method is needed to evaluate the overall dissimilarities between two palm veins feature vectors and check whether or not the degree of likeness between two palm vein feature vectors is higher than the threshold. In general, the threshold is usually calibrated according to the desired security level. In this regard, the higher of the threshold standard, the more intricate of the matching process and the easier for the two palm vein data to be accurately considered as match or no match. On the other hand, the lower of the threshold standard the easier of matching process but the harder for the palm vein data to be considered a match or no match.
In this work we use linear coding theory to solve the problems of securely storing and matching of palm vein feature vectors, since, as motioned above, these vectors are stored in binary forms. Solving this problem directs us to inspect low rate large minimum distance error-correcting codes which should accompany with an efficient decoding algorithm up to the required distance. The author syndrome decoding coset decomposition algorithm introduced in [
To introduce coset decomposition method, we initially need to take into account the basic decoding problem. Let
We say that
Let C be an
In this paper, the preserved palm vein feature vector
where the first k entries of a codeword
We define the syndrome of
Given a feature vector
Here
The palm vein authentication problem often deals with large amount of bits; another difficulty is how to cope with the 10% to 20% of expected errors. Therefore, we require an
Suppose that we have a coding function
As discussed above, the feature vectors of palm vein images acquired from the same person are most probably dissimilar, and needs error detection and correction for matching. Therefore, the matching of the preserved (enrolled) palm vein and the inquest palm vein might be modeled as a noisy channel problem. Once an image of a user’s palm vein is scanned, filters act as a directed smoothing process which removes residual random noise. Then, the extracted feature vector is produced by giving out bits at certain specified positions. Next, the proposed matching algorithm measures similarities between the extracted codeword
The accuracy of any biometric authentication algorithm is measured according to the rates of false acceptance and false rejection. The false acceptance rate (FAR) gives the rate at which someone, other than the actual person is falsely recognized. Thus an unauthorized user is accepted by the system due to erroneous in matching. The false rejection rate (FRR) is how frequently the system fails to match between a palm vein of a person and the same person’s palm vein which is enrolled in the database.
To verify the effectiveness of the algorithm mentioned in section 3, we conducted several experiments using various palm vein images. The algorithm has been implemented in MATLAB and executed on PC with CORE i7 CPU@2.6 GHz, 8G RAM and Windows 8.1 version. Palm vein images (with corrected position and shape) were obtained by a non-contact palm vein sensor that includes an imaging source camera with wavelength of 940 nm. It is worth noting that the sensibility of most security IR cameras falls off between 850 nm and 940 nm. After palm vein image cropping, we transform the normalized image to an 8-bit gray image based on a proposed gray-scale. This is to reduce the complexity of coding problem. By utilizing segmentation process, the extracted feature vector of length m × n × 16, where m × n is the (reduced) size cropped image and 16 is the size of each sub-image, is obtained through extracting LLBP features from each sub-image. Whenever we capture many samples from the same individual, the values of each bit in same location of LLBP codes should be all identical. This is the ideal situation, however, some bits may have different values as a result of considering the interference factors such as displacement and rotation. There are also small gestural differences or changes in the palm vein over the time. As a result of all these negative effects in recognition and matching performance, we use the error detection and correction (coset decomposition) algorithm. Therefore, LDPC is used first to encode binary feature vector and second to measure the similarity between the extracted and enrolled (registered in database) binary codes.
The database from 50 users, with five images per user (with minimum interval of one week per each image), was acquired and used as a basis for the system database. A randomly selected image for each user is used to test the proposed approach. It has found to report higher verification accuracy of 99.8% for FAR and 99.6% for FRR.As expected the identification performance improves as the length of feature vectors increase. The performance of the palm vein authentication algorithm is shown in
Palm vein pattern recognition is an upcoming biometric technology which is easy to train, use, implement and operate. Furthermore, it is a convenient technology with high security and accuracy level. It can be used for personal computers log-in, building entrance and ATM machine. It can also be used as identification verification for medical record management. In spite of the fact that the palm vein authentication is gaining momentum, it is still untested as it has not yet been available for mass production. There are some factors that might affect the quality of the captured images such as user body temperature, ambient temperature, humidity, heat unevenly distribution, heat radiation, camera calibration and focus of camera. In addition, this technology is not applicable to people who lost their palms.
We have developed an algorithm which merges the biometrics and cryptography. Some of biometric matching algorithms analogous to the presented algorithm have been proposed in the literature. Our model which is fit for matching user palm veins is, to the best of my knowledge, entirely new. We can apply the same approach to different biometrics systems, for instance, the iris biometrics. In particular, the iris matching could be easier in terms of high true match rates because of the large amount of information that might be extracted from an iris. A possible extension to the described approach would be, to some outlines, to build a multi-biometrics authentication system.