Journal of Electrical and Electronic Engineering

2021; 9(5): 147-152

[Link]
doi: 10.11648/[Link].20210905.11
ISSN: 2329-1613 (Print); ISSN: 2329-1605 (Online)

Fingerprint Classification Using Deep Convolutional Neural

Network
Mohamed Hirsi Mohamed
Department of Electrical and Electronics Engineering, Cyprus International University, North Cyprus, Turkey

Email address:

To cite this article:

Mohamed Hirsi Mohamed. Fingerprint Classification Using Deep Convolutional Neural Network. Journal of Electrical and Electronic
Engineering. Vol. 9, No. 5, 2021, pp. 147-152. doi: 10.11648/[Link].20210905.11

Received: August 5, 2021; Accepted: August 16, 2021; Published: September 6, 2021

Abstract: Fingerprint classification is a method of reducing the number of candidates needed by fingerprint recognition
systems to determine if a fingerprint picture matches one in the database. Deep learning has gained a lot of attraction in the
recent decade including natural language processing, digital image processing, speech recognition, handwritten digit
recognition, medical picture assessments, and so on. The subject of this paper is to explore the factors affecting fingerprint
classification using a convolutional neural network and to train and test a deep CNN model, The CNN model includes two
serial stages, a preprocessing phase which is used to enhance the fingerprint images qualities, and post-processing phase which
used to train the classification model. This has been accomplished by designing a new deep convolutional neural network
model for this work. The Convolutional neural network model achieved outstanding classification accuracy on the fingerprint.
This experiment used the NIST DB4 dataset which contains 4,000 fingerprints images with five labels. Separately, each label
of this database comprises almost 800 fingerprint samples with dimension of 512 x 512. To lower the training time required we
reduced the fingerprint images up to 200 x 200 dimension. the study achieves 99.2% of classification accuracy with a zero-
rejection rate.

Keywords: Biometric, Fingerprint, Convolutional Neural Network, Overfitting, Classification

integrated automated fingerprint identification system. Low

1. Introduction enforcement agency computers respond at least half house for
Biometrics is used to measure and discovering human any identification request [1]. To minimize the time consuming
identity based on physical, behavioral, and chemical of the fingerprint recognition system, first, we need to identify
characteristics. The Biometric system can be fingerprints, fingerprint class. Identifying the fingerprint classification is not
odor, face, and voice. Fingerprint and face can be characters define the personal identity although the perfect classification
as physical traits, voice is behavioral characteristics, and odor can shorten the authentication process by decreasing waiting
is a chemical attribute. time, searching, and pointless computation. The Federal
A fingerprint recognition system is an automated method Bureau of Investigation itself mentions in its NBIS print
that describes a personal identity, based on a process of program guideline that “To know the class of the fingerprint is
comparing stored database with the input fingerprint image. effectively decreasing the amount of candidate that searching
Every human being has a unique fingerprint at any age and fingerprint need to decide if the print matches the database
they are patterns that never change it is appearance even files” [2, 3].
when it is burned it will grow back with it is an original Normally, fingerprints can be classified into three main
pattern. Fingerprints are the most common and well know classes; arches loops and whorls, and then increase up to five
authentication system. classifications by separating Loops into two categories (left
Fingerprint biometric systems are very common tasks in the loop and right loop), and arches into two categories (tented
forensic lab and law enforcement teams. in the last decades, and arch).
the United States of America police and the Federal Bureau of Fingerprint classes are not consistent, both whorl and Loop
investigation FBI stored millions of fingerprint images in an fingerprints are the most common fingerprint with almost
148 Mohamed Hirsi Mohamed: Fingerprint Classification Using Deep Convolutional Neural Network

94% of all fingerprints exist. only 6% of human being has Oxford’s VGG architecture model was used to applied
arch patterns. classifications. The authors implement three different
This paper is arranged as follows: part II related work, part models; First architecture has the shortest learning time and
III Methodology, part IV Result, and lastly part V conclusion. lower accuracy, in the next model the accuracy has increased,
and in the last model experiment has achieved 97.1%.
Another study conducted by Hadi Veisi and Behnam
Bakhshi with the title of “End to End Fingerprint Verification
Based on Convolutional Neural Network” was proposed two
CNN architectures [5]. FVC 2002 dataset was used as an
experiment dataset with augmentation techniques to escape
overfitting issues. This dataset comprises 110 fingerprint
images. Ten images were used for testing, 10 fingers are used
for validation, and the rest 90 fingers were used for the
training phase. The pre-trained AlexNet model is used to rise
training speed and original convolutional neural network
which contains both classification and hidden layers. The
proposed model has no prepossessing requirement. The
authors accomplished an equal error rate of 17.5%. this is an
improved consequence compared to A-KAZE which
provided an EER of 41.6%.
Paralt, et al. use a convolutional neural network to classify
finger patterns [11]. In this study, CaffeNet pre-trained were
implemented and proposed a new CNN model for fingerprint
pattern classifications. The new proposed model comprised
multiple layers such as a convolutional layer, fully connected
layer, and pooling layers. In Both model’s stochastic gradient
descent were trained. NIST-DB4-F and NIST DB4-S datasets
were used to train and test the model. The authors of [11]
Figure 1. Fingerprint classes: (a) loop, (b) Arch, and (C) Whorl. achieved different accuracy rates, on NIST-DB4-F 90.7%
accuracy was achieved, while NIST-DB4-S achieved 88.9%
2. Related Work with zero rejection rate. all experiments were conducted on a
core i7 computer with 24GB Random access memory with
A research study conducted by Shrein [3] with the topic Nvidia Graphic Processing Unit. This experiment has no
of “Fingerprint Classification Using Convolutional Neural preprocessing.
Networks and Ridge Orientation Images” proposed
Convolutional neural network architecture for fingerprint
classification. The author uses NIST-DB4 which
3. Methodology
comprises 1650 image samples. Gaussian filter is used to
smooth fingerprint images, then ROI was selected. Finally,
data augmentation was used to generate an extra image
from the existing database. In this study CNN architecture
was used as postprocessing, the architecture contains three
different layers; pooling layers, convolutional layer, and
fully connected layers. All layers were used ReLu for
activation function to decrease the unsupervised pre-
training in the CNN. Shrein conducted all his experiments
on laptop core i7 with sixteen gigabytes of memory and an
Nvidia graphics processing unit. Experiments use 50
epochs and took almost 20 seconds for each epoch. Finally,
the study achieved an accuracy rate of 93.5% for five
labels and four labels 95%.
Yong Rhee and his research fellow studied fingerprint
pattern classification using a convolutional neural network
[4]. Use fingerprint verification competition 2000, 2002, and
2004 dataset in five labels. To enhance fingerprint image
quality normalizing and segmentation process was applied, Figure 2. Illustration of model Methodology.
then binarization and Gabor filter were used. In this paper,
 Journal of Electrical and Electronic Engineering 2021; 9(5): 147-152 149

3.1. Preprocessing 3.1.2. ROI Segmentation

It is useful to know the region of interest (ROI) of all
The convolutional neural network model has a bunch of fingerprint matters. As it only contains background
parameters that could be learned in the train section. To train information, the fingerprint sections that are not ridge or
the CNN network, and to achieve perfect performance, we valleys were ignored. The latest limits of the active area are
need a large dataset with detailed classes. Though Currently, displayed when the details of the limit area are confusing the
several fingerprint databases are accessible online, it is very fine details created when the sensor is outside. Two
difficult to choose a database with enough capacity to train techniques are used to find areas of interest ROI.
the CNN model. To assess and choose a public database we a. Block direction guesstimate and direction variety
are looking at the quality of the image, number samples b. Morphological Technique
dataset contains, and class labels. due to its quality, ROI is used to remove the background part of the
NISTDB4 was preferred to implement this research. The fingerprint. However, the convolutional neuronal tests and
database has 4,000 samples and five different category labels, training will be equal or more effective without using ROI.
and it's free. There are only 2000 unique fingerprints in
NISTDB4, which means that each fingerprint is sampled 3.1.3. Binarization
twice [6]. Because more images are more suitable for training The process of converting gray-level images into a black
and testing CNN networks. and white image is known as binarization. basically,
reduction the image information from 256 shades value into
Table 1. Database sample categories. two black and white binary images. This process is also
Class name Symbol Samples known as image threshold processing. The images are
Arch A 800 samples divided into component objects and are generally
Left Loop L 800 samples implemented to extract or define objects of images [7].
Right Loop R 800 samples
Recently, there are several techniques used for image
Tent T 800 samples
Whorl W 800 samples binarization, but there is not one single technique to apply all
Total of samples 4,000 samples the image.
1 ,
3.1.1. Ridge Orientation g x, y (1)
In general, fingerprint patterns comprise local features such 0 ,
as co-directional regions which know as minutia, and global
features. Local ridges are useful for individual identification or
verification. In global function, the orientation of the ridge is a
functioning of obtaining the ridge angle of all fingerprint
images. In almost all fingerprint procedures, the value of the
ridge line of the fingerprint brings severe effects [7, 8]. You
can use orientation fields for the improvement of the quality of
the fingerprint. Gradient-based approaches are the most
common technologies used to calculate Ridge orientation. In
addition, high performance can be achieved using a Gradient
method during the CNN test.

Figure 4. Binarization result.

3.1.4. Thinning
Thinning fingerprint patterns is a process of reducing the
thickness of the ridge line in image pixels by removing image
pixels at the corners of the image ridge. Generally, the basic
concept of thinning a fingerprint image is to generate a
template and then match the binary image to the created
Figure 3. Ridge Orientation. template to confirm whether a pixel is recognized at a certain
point. This reduction can be achieved in the skeleton of the
150 Mohamed Hirsi Mohamed: Fingerprint Classification Using Deep Convolutional Neural Network

entire region by thinning techniques. Figure 5 shows the and the number of network outputs is the same.
result of thinning.

Figure 5. Fingerprint Thinning with crop selected.

3.2. Network Model Architecture

CNN has been proven to be one of the active promising

algorithms in image classification tasks, including classic
problems such as Fingerprint classification, handwritten
recognition as well as others [9, 10, 12, 13]. It is offering a
real-life solution to other fields such as motion synthesis,
facial recognition, and pose estimation. For training and
testing deep convolutional neural network models, every Figure 6. Illustration of convolutional model architecture for this work.
input image will pass through over a series of layers. The
layers are defined as follow: The first FC layer and all convolutional layers use ReLu as
a. Input data: all input image sizes in this experiment the activation function to cut unsupervised pre-training in
should be 200x200 dimensions. CNN layers. All experiments are performed using Matlab,
b. Convolutional layers: every conv layer consists of a which offers a deep learning toolbox. The architecture model
bank of kernels which implemented with the output of of this study is shown in Figure 6.
the earlier layer, or the input data if it is not the first
convolution layer. Here in this study, the first two layers
are a convolutional layer with 32 kernels, 5x5
4. Result
dimensions, and a stride of 3. NIST database was divided into the training portion and
c. Pooling Layers: in this work max-pooling was used as it testing portion. 20% of the data were used to test the
is a perfect technique to perform downsampling on the convolutional neural network model. A group of fingerprints
output data by selecting the maximum value. Here images was used to input on the network model one by one.
Max-pooling has 2x2 dimension with 2 strides. every attempt of loading one fingerprint, a loop function was
d. Another two convolutional layers were added with 64 applied to confirm if the network was classified correctly.
kernels of dimension 2x2. The class categories in the database class have an
e. Dropouts: are used to cut the weight connection appreciation such as R which stands for Right loop, A which
between layers. 0.2 dropout was implemented to means Arch, L for Left loop, T present for Tent and W for
remove the 20% weight between the FC layer and the Whorl fingerprint.
convolutional layer. Table 2 demonstrates that the trained algorithm has trouble
f. Fully connected layer: these layers perform a flattening classifying the Tents fingerprint pattern, with a 1.75 percent
process to convert the image matrix into a dimensional misclassification rate. The Tents fingerprint pattern resembles
vector. The Fully Connected layer contains 128 neurons. the Arch fingerprint pattern quite closely. The similarities
Finally, there are 5 neurons attached to a single-layer FC, among both patterns are that both ridges begin on one side of
 Journal of Electrical and Electronic Engineering 2021; 9(5): 147-152 151

the fingerprint and flow to the opposite side. the ridges of the second group is testing, which makes up the remaining 20%.
Arch fingerprint pattern move from one side and irritated to During training, we see that the number of fingerprints in the
the other side as a waveform in the center of the pattern. The database is growing. Monitoring the quality of the training
arch pattern's ridge is not continuous, as it is in a Tent pattern, database and choosing different training parameters can help
but rather lies in the middle of the fingerprint pattern. increase the model achievement rate.
The assessment results of the deep convolutional neural
Table 2. Confusion matrix using the Proposed CNN fingerprint classification network model training and validation classification accuracy
Predicted.
rate over the 10 epochs are shown in Figure 7. Classification
True class Arch Left loop Right loop Tent Whorl accuracy rate is described as the number of valid predict to the
Arch 791 0 0 4 0 total number of predict made. The proposed CNN model
Left loop 0 797 0 4 0
network has achieved by 99.2% of classification accuracy rate.
Right loop 1 0 797 6 1
Tent 8 3 3 786 0 On the other side, we have less false classification and the
Whorl 0 0 0 0 799 highest classified rate, with five labels of fingerprint classes
Accuracy = 99.25% being used to train a CNN model and determining the
number of neurons required in the model's output layer. The
As the dataset was split into two groups. The first group is model also ignores the issue of overfitting, which means less
training, which makes up 80% of the database, and the complexity of the system.

Figure 7. Illustration of Model accuracy rate using 10 epochs.

Table 3. Comparison between the proposed model and related works Method.

Method Data Number of Samples Classified Numbers Accuracy

Proposed CNN Model NISR-DB4 4,000 5 99.2
VGG [8] NIST-DB4 4,000 4 95.5%
CaffeNet pre-trained [11] NIST-DB4-S 3,300 5 88.9%
CNN [3] NIST-DB4 4,000 5 95%
Pre-trained AlexNet [5] FCV 2002 110 - 82.5%

orientation, and Binarizing were used to enhance the

5. Conclusion fingerprint images.
The main advantage of the proposed network is the very
In a biometric system, the human fingerprint includes high accuracy rate. Similarly, the network model can execute
patterns that may be used as a secret password. In this automatic fingerprint pattern classification without a human
study, the basics of human fingerprint classification are interacting. in this respect, when the input image is uploaded,
discussed. Besides, a new CNN architecture model was the preprocessing stage will be applied and then to the trained
proposed and tested to classify fingerprints. The model a CNN layer will implement. The fingerprint feature is
consists of two main processes, the post-processing step extracted along with the kernels learned during the training
which is used to train and classified the fingerprint image, period and lastly the class label of the fingerprint is
and preprocessing stage for increasing the fingerprint classified.
quality. Thinning techniques, ROI segmentation, ridge
152 Mohamed Hirsi Mohamed: Fingerprint Classification Using Deep Convolutional Neural Network

Grey Scale Images", International Journal of Computer

Applications, vol. 71, no. 1, pp. 8-11, 2013. Available:
References 10.5120/12320-8533 [Accessed 30 June 2020].

[1] FBI, (2014). Integrated Automated Fingerprint Identification [8] Michelsanti, D., Ene, A-D., Guichi, Y., Stef, R., Nasrollahi,
System. Federal Bureau of investigation. Available at: K., & Moeslund, T. B. (2017). Fast Fingerprint Classification
[Link] with Deep Neural Networks. In F. Imai, A. Tremeau, & J.
management/foipa/privacy-impact-assessments/iafis Braz (Eds.), VISiGRAPP 2017: Proceedings of the 12th
[Accessed 19 Feb. 2020]. International Joint Conference on Computer Vision, Imaging
and Computer Graphics Theory and Applications (Vol. 5, pp.
[2] Wilson, C., Candela, G., and Watson, C. (1994). Neural 202-209). SCITEPRESS Digital Library.
network fingerprint classification. Journal of Artificial Neural [Link]
Networks 1.2, pp. 203-228.
[9] Zhong Z, Zheng L, Kang G, Li S, Yang Y (2017) Random
[3] J. M. Shrein, "Fingerprint classification using convolutional erasing data augmentation. arXiv. Available online at:
neural networks and ridge orientation images," 2017 IEEE [Link] [Link]. [Accessed 16 Feb
Symposium Series on Computational Intelligence (SSCI), 2020].
Honolulu, HI, 2017, pp. 1-8.
[10] M. Galar et al., "A survey of fingerprint classification Part II:
[4] W. Jeon and S. Rhee, "Fingerprint Pattern Classification Using Experimental analysis and ensemble proposal", Knowledge-
Convolution Neural Network", international journal of fuzzy Based Systems, vol. 81, pp. 98-116, 2015. Available:
logic and intelligent systems, vol. 17, no. 3, pp. 170-176, 10.1016/[Link].2015.02.015 [Accessed 1 March 2020].
2017. Available: 10.5391/ijfis.2017.17.3.170.
[11] Peralta, D., Triguero, I., Carcia, S., Saeys, Y., Benirez J. and
[5] B. Bakhshi and H. Veisi, "End to End Fingerprint Verification Herrera, F. (2017). Robust classification of different
Based on Convolutional Neural Network", in 27th Iranian fingerprint copies with deep neural networks for database
Conference on Electrical Engineering (ICEE2019), Tehran, penetration rate reduction.
2019, pp. 1994 - 1998.
[12] Yaniv Taigman, et al. (2014). DeepFace: Closing the Gap to
[6] C. Watson, "NIST Special Database 4. NIST 8-Bit Gray Scale Human-Level Performance in Face Verification. Conference
Images of Fingerprint Image Groups", NIST, 2020. [Online]. on Computer Vision and Pattern Recognition (CVPR).
Available: [Link]
database-4-nist-8-bit-gray-scale-images-fingerprint-image- [13] Guilhem Cheron, Ivan Laptev, Cordelia Schmid. (2015). P-
groups. [Accessed: 01- Nov- 2019]. CNN: Pose-Based CNN Features for Action Recognition.
Proceedings of the IEEE International Conference on
[7] P. Puneet and N. Garg, "Binarization Techniques used for Computer Vision (ICCV), 2015, pp. 3218-3226.