Cite this as
Ashraf S, Ahmed T (2020) Dual-nature biometric recognition epitome. Trends Comput Sci Inf Technol 5(1): 008-014. DOI: 10.17352/tcsit.000012All humans are born with unique physically identified body characteristics to other persons which remains unchanged throughout life. These characteristics are taken into account by the emerging technology to get recognized from person to person. The technology used by the traditional human identification system sometimes becomes inefficient when data or images received are not up to the acceptable quality mark or when a person has a face covered with mask-like during epidemic virus fistula. In order to overcome such human recognition challenges, a Multi-biometric sustainable approach (D-nb) has been proposed using two uniquely identified modalities like foot and iris. This approach shrewdly identifies and makes the bodacious recognition among humans and suggests the sagacious result which is foremost better than the traditional biometric system.
Today a biometric recognition process is very popular from both physiological and behavioral points of view for identification and authentication of individuals. If both are the same as the decision, the test picture corresponds to the store picture and it is checked that it is a mapping operation [1]. In order to identify the test image with the saved database, the person is identified or does not decide whether it is only one to many mapping systems. Some prominent biometric identification features are being used as [2].
Face biometric: Face recognition is a technique in which a person is automatic identified and verified using digital image [3].
Ear biometrics: Ear biometrics uses features or characteristics of ear for matching. This is a stable and not changes with age [4].
Fingerprint biometrics: A fingerprint is one of the most popular biometrics used in last few decade for recognition of a person. Ink is used for taking an impression of finger and thumb images on paper [5].
Hand Geometry biometrics: Now palm print is also used as a biometric because each person has no changes in lines of palm print during the growth [6].
Voice biometrics: Voice biometric comes under the behavioral characteristics each person has different sound, pitch and speaking style [7].
Signature biometrics: The making of signature. Signature are measured as writing style of a person that means holding the pen, pressure on paper and the direction of letter is different each parson has its own style [8].
It accomplishes in two steps; one is the training process and the other is the testing process.
o Input image is captured or acquired by camera.
o An image is pre-processed means image resizing, reformatting.
o Feature extraction is performed.
o Prepared database.
Testing of image: It is close to the preparation phase, but balanced. The processes of training and testing are described in Figure 1
There are numerous challenges faced by the Uni-biometric system such as:
• Noise
• Intra-class variations
• Non-universality
• Spoof attacks
Multi-biometrics offer opportunity to take more than two biometric features, to combine them to overcome unimodal biometric system limitations and to achieve optimal results [9]. We’ve been using two biometric foot and iris for a multi-biometric device. Sprints have been developed, iris are special and not changed during a person’s age as a reflective layer structure.
A critical comparative analysis of foot and iris modalities with other available practices have been epitome in Table 1.
The rest of the article is arranged as, Related work is addressed in section II, section III presents the proposed methodology, section IV demonstrates the performance obtained through results while concluding remarks are given in section V.
Many researchers have made significant contributions reported [10], that the biometrics is becoming more commonly used in several places through different biometric device such as bank, institute, airports. Multi-biometric is preferred in authentication-based image processing applications. The authors [11], introduced a multi modal systems working on physiological and demography information for identification of person.
The author [12], links open overlay panel introduced a biometric system for person recognition that shown the improvement accuracy of the recognition system. Another research work on face and ear-based hybrid system has been presented by [13]. This system is used Iterated Function concept in which images are compression and indexed. Sarfraz, et al. [14], designed a multi-biometric system that combines face modalities and foot modalities using PCA [15], classifier for face and wavelet transformer for foot and concatenated after normalization process to obtain a matching score and take a decision. Another author Davrondzhon [16], proposed a system for multimodalities biometric that combines face, ear and iris using PCA, Eigen image [17] and hamming travers [18], classifier for feature extraction, sum rule method is used for fusion to calculate matching score. They introduced recognizing person utilizing multiple biometric traits and their advantages such as high accuracy and robustness, that increased recognition performance. Herbadji, et al. [19], developed a Multi-biometric system, which combined face and foot modalities at decision level using PCA classifier for face and wavelet transformer for foot and concatenated after normalization process to obtain matching score [20].
The body parts can be paired with the other body elements in several potential ways, leading to very diverse machine architectures. Particularly in parallel or in cascade, the Iris footprint can be combined and therefore further findings are arranged as:
Iris recognition work on the colored part of eye that is isometrics surrounds by pupil known as iris. Iris image is captured using high quality camera every person has a unique iris. Process of calculating patterns are very complex it stored in bit format as a template illustrated in Figure 2. Hamming traverse method is used for calculating the distance between test image and stored template if the distance is zero that indicate the test image and training image is same otherwise, we can say both images are different.
Identification of peoples by fingerprint is becoming a wonderful success. Measurement such features are not a complicated process some transformation such as furrier, haar are available [21]. Each parson has completely different footprint and it is easy to capturing without any special requirement both leg images can be used for recognition [22]. Captured foot image needs some additional approach such as cropping and resizing. Figure 3 displays the picture captured by high quality camera than the RGB image converted into gray scale after that resize them for preparing the database.
Hamming distance approach [23], is applied for a matching iris templates incorporate noise masking also. The distance between two iris templates is calculated in bits format and distance calculated between two iris templates if the distance is minimum that indicate the iris are same otherwise iris are not matched which confirms hamming distance is a reliable and accurate for iris recognition.
The iris recognition process completed in three stages
a. Segmentation: Located the iris region
b. Normalization: Created a consistency of data iris region
c. Feature encoding: Produced templates as set of discriminated features of the iris.
The sequential modified haar wavelet [24], is based on the mapping technique in which integer-valued signals are mapped using the reconstruction of image technique. Wavelet coefficients represented by decimal numbers and each decimal value needed eight bytes for storing the haar value.
The sample images of eye are being consider as input to the system and thereby iris templates are generated as output. The discrete bits formation of the image has been obtained illustrated in Figure 4, showing segmentation and normalization stages [25].
The control area on the 3-D floor is exposed to the pressure distribution of the person on the medium [26]. Because human physical presence defines the significant pressure surfaces, personality identification can be thus understood through the classification of high pressure surfaces and examination of their formal features [27]. When the scene footprint is compared to other templates, the public protection program can assess or reject the accused individual. The visible footprint arc travels away from the curved surface position and direction which has little to do with the curved surface parameter.
For identification of footprint patterns, edge detection is necessary. For the sake of the gradient of light intensity of a gray-scaled picture it is very difficult to distinguish corners via a Laplacian transformation [28]. Then for edge detection the entire picture must be binarized. To binarize the entire image, a certain threshold value is not adequate. The entire picture is then separated into several sections, and threshold values are autonomously determined for such a segment.
The proposed technique produces n number of iris codes out of n number of eyes tests performed at various time periods from the same person. Using majority voting scheme [29], a unique iris code x is built from the n number of iris code. The suggested framework operates on x and generates codewords, known as Error Corrected Iris Code (ECIC) [30]. These ECICs are composed of iris coding and parity regulation p. The hash h is also created from code x. Eventually, parity regulates p of the ECIC, matrix h of parity and hash-h render code that are embedded in digital images by transformation of the integrated wavelet and incorporation processing threshold illustrated in Figure 5.
The new and unique feature of the multi-biometric is the speed and precise segmentation of irises, the iris image being captured to reduce identifying errors, producing a vector with distinguishing texture characteristics and a proper dimension to improve the detection and computational efficiency [31]. The sagacious edge detection and circular Hough transforms are used for the segmentation process. The segmented iris is normalized and phase data is extracted by applying Log-Grabor filter [32] and encoded shrewdly to produce the refined vector feature through phase quantization method [33]. The iris from real eye object has been extracted after undergoing segmentation, normalization and quantization process as shown in Figure 6. Unless the iris area has been effectively segmented from the eye object, the next step is to turn the iris region in such a manner that it has set measurements to enable comparisons [34]. Dimensional incoherence [35], between eye images are mainly because of the iris stretching caused by the dilation of pupils from different luminance levels. Other sources of inconsistency include varying imaging distance, camera rotation, head tilt and eye rotation within the socket of the eye. Construction outcomes of individual iris code x from n number of iris code are achieved using a specific procedure named extraction vote. Figure 7, explains the fabrication of these unique iris x from three sample iris codes. The sparse parity test matrix is used to
describe multi-biometric codes. This small matrix is often generated at random, depending on the sparsity limitations [36]. Referring Figure 5, The valid message bits are processed by the multi-biometric phase when satisfy the condition thereby extraction voting occurs. After creating a specific iris code x, each column in iris code x is considered to be a message in multi-biometric encoding and encoded to render ECIC with the aid of generator values. The special iris code x is converted into hash h, often referred to as cancellable iris code. Table 2 shows factual score, imposter score and threshold value for object 1 and 2 respectively. The Iiris and foot biometrics were
tested individually and Table 3 show False Accept Rate (FAR) [4], and False Reject Rate (FRR) [37], of Iris and foot. The weight of all individual iris and foot modalities was calculated, shown in Table 4. Iris recognition and foot recognition algorithms produced dissimilarity scores [38].
The Min-Max normalization technique [39], was used to convert all dissimilar data into similar data shown in Table 5.
By adopting weighted sum rule procedure, the matching scores generated for both modalities are fused and combined matching scores are generated as Equation 1.
Considering S as a final score obtained by averaging and then adding the scores of iris (I) and foot (F). If the results match successfully, person will eventually be verified, otherwise, depending on the matching score.
Table 6 shows the weight value of iris and foot after the fusion of two traits.