Volume 8, Issue 2, February – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Deep Gender Identification Model with Biometric

Fingerprint Data
Diptadip Maiti Madhuchhanda Basak
Computer Science & Engineering Dept, Computer Science & Engineering Dept,
Techno India University, Techno India University,
Saltlake, Kolkata, West Bengal Saltlake, Kolkata, West Bengal

Abstract:- People may be easily distinguished from one deep convolution network which can accomplished
another thanks to their distinctive and special traits, which classification of gender [8] [9] from fingerprint with competing
also serve as a means of identification. One of the most with acceptable outcomes. The manuscript is organized as
important pieces of identification information is gender. If follows: Section I, following the introduction, related work in
we can confidently determine a person’s gender, it will section II and proposed methodology in Section III. The
reduce the number of inquiries and shorten the search outcome of the proposed method is discussed in Section IV and
period while increasing the likelihood that someone will be remaining conclusion and future scope in Section V and
recognized. In this work, we apply deep convolution Section IV respectively.
Neural Network to classify fingerprints by means of
gender. The proposed model achieves an validation II. RELATED WORK
accuracy of 96.46% for the classification of gender.
Publicly available Sokoto Coventry Fingerprint Dataset In the recent past, there has been a lot of study into
(SOCOFing) is applied as a benchmark for the outcome of fingerprint identification and classification. But relatively few
the classification accuracy of the proposed network. studies have looked into the issue of gender categorization and
identification using fingerprints. Ahmed Badawi et. al. [6]
Keywords:- Biometric, Fingerprint, Deep Learning, devised a Gender categorization from fingerprints, which is a
CNN,Gender Identification,) crucial step in forensic anthropology. An analysis was done on
a collection of 10 fingerprint pictures representing 2200
I. INTRODUCTION people of various ages and genders (1100 men and 1100
females). Ridge count, ridge thickness to valley thickness
Biometric data includes a person’s voice, signature, ratio.
handwriting, iris, fingerprint, face, and other measurable,
distinguishing physical characteristics. Biometric data is (RTVTR), white lines count, ridge count asymmetries,
gaining popularity as a security tool since it is unique to each and pattern type concordance were among the features
individual and offers a high level of security. One collected. To classify data based on the most prevalent
distinguishing feature of the fingerprint is that each individual’s characteristics, Fuzzy - C Means (FCM), Linear Discriminant
form is unique and does not change over the course of a lifetime Analysis (LDA), and Neural Network (NN) were employed.
[1] [2]. A fingerprint may therefore be used to precisely identify They used FCM, LDA, and NN to get outcomes of 80.39%,
and confirm a person, and it has recently become used for 85.5%, and 88.5%, respectively. Ashish Mishra et. al. [10]
mobile devices like smart phones as a rapid identification successfully reduce search time by managing the searching
method [3]. The interpretation of fingerprints is dependent on stage for both right hand and left-hand thumb datasets using
the Locard’s Principle of Exchange. The discharges found in the classifiers SVM and Naive Bayes. Prabha et. al. [11]
fingerprints include traces of different chemicals and their proposed a research technique of gender categorization using
metabolites, which can be discovered and used for forensic fingerprints which is based on characteristics that were
analysis. They can be found at the crime scene, making it easy retrieved using Discrete Wavelet Transform. S.S. Gornale et.
to confirm the presence of a suspect, victim, or anybody else. al. [7] put into practice a cutting-edge method for gender
These days, fingerprints are routinely used in workplaces and detection using fingerprint pictures. KNN and LBP classifiers
educational establishments to verify a person’s existence [4] are employed. The gender detection percentage for a collection
[5]. Although the human vision system is self-sufficient and of male and female fingerprint pictures using this method is
quite adaptive in roughly recognizing someone’s age and 95.88%. A method developed by S. Falohun et. al. [12] uses
gender, it is still difficult or impossible for the machine to tell fingerprint analysis to determine a person’s age and gender.
whether a male or female fingerprint belongs on a person. As a Machine is trained using integrated DWT & PCA
result, gender identification [6] [7] based on any biometrics characteristics (for age classification). Ridge / Valley count
bisects the search space. with Back Propagation Neural Network is used to classify the
gender using Ridge Thickness Valley Thickness Ratio
Only a small number of academics have studied characteristics. Mangesh Shinde et. al. [13] used SVD and
fingerprints for gender categorization and unearthed the best DWT to identify a person’s gender by utilizing fingerprint. In
outcomes. In this proposed work, we are trying to develop a the research, the fingerprint image is examined up to six levels

IJISRT23FEB1204 www.ijisrt.com 1485

Volume 8, Issue 2, February – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
of DWT and SVD separately, followed by the combined result B. Workflow:
vector of DWT and SVD. Overall accuracy rates for both male The fingerprint images from SOCOFing dataset is first
and female participants in the trial were 78.65%. Rekha et. al. preprocessed and resized to 96x96 for the input of the
[14] uses Gabor filter and Support Vector Machine for gender proposed deep learning model. The detail work flow of the
classification from fingerprint. The Adaptive Nero-Fuzzy proposed method is shown in the Fig 2.
Inference System (ANFIS) approach is used by Suman Sahu
et. al. [15] to identify fingerprint gender. Frequency Domain
Analysis (Frequency-based properties including vertical,
horizontal, amplitude, and diagonal) and RVA must be
extracted in order to group fingerprint images as female or
male. Terh¨orst et. al. [16] demonstrated that a single minutia’s
gender choice accuracy is linearly correlated with its
dependability, and that adding this information in the fusion
process improves performance. A feature vector made up of
the ridge thickness to valley thickness ratio (RTVTR) and the
ridge density values was used by Sarath et. al. [17] to represent
each fingerprint in the database. They achieve a reliable
classifying function for male and female feature vector
patterns using a support vector machine trained on a batch of
150 male and 125 female fingerprints. Merkel et. al. [18] uses
the capturing devices (CWL and CLSM), to capture
fingerprint images, that covers ten dissimilar datasets with
entire 2,618 time-series. Correlation coefficients and age
estimate based on machine learning are the supplied baseline
findings for age estimation on all 10 sets. Shivanand Gornale
et. al. [19] employed Gabor and DWT-based characteristics to
classify fingerprints into male and female categories. The Fig. 2. Proposed methods’ Workflow
fingerprint pictures of 74 individuals, representing a range of
ages and genders, were acquired for the system’s training. C. Preprocessing:
Prior to training and testing, the 96x103 pictures in the
III. PROPOSED METHOD dataset are downsized to 96x96 and given a grey label value of
0-1. Three sets have been created from the dataset. A total of
A. DataSet: 14781 photos were utilized for testing, 10347 images for
Training and testing of the proposed method, was carried validation, and 24142 images for training.
out using publicly available dataset named Sokoto Coventry
Fingerprint Dataset (SOCOFing) [20]. The dataset contains D. Network Architecture:
6000 real fingerprint images which again treated with A four-convolution layer is used to extract features from
obliteration, central rotation, and z-cut with different degree to the input fingerprint, which is then followed by batch
form altered easy, altered medium and altered hard fingerprints normalization and maxpooling. The final extracted feature is
images. The size and other details of the dataset is shown in the obtained by doing a global maxpooling operation following
Fig 1. the convolution procedure. The extracted feature is passed to
a artificial neural network containing 128 hidden node which
classify the feature in to required output of male / female. Fig
3 displays the network architecture of the proposed method.

E. Network Hyperparameter:
Activation function used in convolution layer is rectified
linear unit and he uniform is used as initial network weight
assignment. Softmax activation function is used in the fully
connected dense layer of the artificial neural network
classifier. For optimization of the performance of the network
Adam optimizer is used in the proposed network. Finally
categorical cross entropy is used as the loss function of the
network, which is minimized by the proposed network. Fig 4
depicts the specifics of the proposed network.

Fig. 1. Details of Dataset SOKOFIG F. Network Training:

With a batch size of 64 the network was trained for 100
epochs to achieve the desired accuracy of the classification of
gender from fingerprint. The network takes approximated 12
hours for training the part of whole dataset used for training.

IJISRT23FEB1204 www.ijisrt.com 1486

Volume 8, Issue 2, February – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig. 4. Proposed Network Parameter Details

IV. EXPERIMENTAL RESULTS AND

PERFORMANCE ANALYSIS

We conducted extensive experiments using deep

convolution networks to demonstrate the usefulness of our
suggested technique. Table I shows the accuracy, loss value of
training and validation for gender identification. In Table II the
proposed methodology has been compared with other
methodology from the related current work in view of
classification performance for gender identification for one
batch. The classification report of the proposed network is
conveyed in the Table III. Fig 5, Fig 6, Fig 7 and Fig 8 displays
the training and validation accuracy, training and validation
loss, Precession and recall with receiver operating
characteristic curve of the proposed model during training and
testing.

Table 1 Training & Validation Accuracy And Loss For

Gender Classification
Gender Accuracy Loss Value
Identification
Training 99.86% 0.0034
Validation 99.33% 0.0266

Fig. 3. Proposed Network Architecture

IJISRT23FEB1204 www.ijisrt.com 1487

Volume 8, Issue 2, February – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Table 2 Comparison of Proposed Method with Current Work.
Sl No. Title Method Used Accuracy
1 Badawi et. al.[6] FCM, LDA, NN 80.39%, 85.5%, 88.5%
2 Sheetlani et. al.[11] BPNN 96.6%
3 S.S. Gornale et. al.[7] LDA & QDA 95.8%
4 A. S. Falohun et. al.[12] ANN 80%
5 S.S. Gornale et. al.[19] KNN, LBP 95.8%
6 Mangesh et. al.[13] SDV, DWT 78.6%
7 Proposed Method CNN 99.33%
TABLE III
CLASSIFICATION REPORT FOR GENDER
Precision Recall F1-Score Support
Male 0.99 0.99 0.99 111
Female 0.94 0.94 0.94 17
Accuracy 0.98 128
Macro Avg 0.97 0.97 0.97 128
Weighted Avg 0.98 0.98 0.98 128

Fig. 5. Training and Validation Accuracy Fig. 7. Precession & Recall of the Network

Fig. 6. Training and Validation Loss Fig. 8. ROC Curve of the Network

IJISRT23FEB1204 www.ijisrt.com 1488

Volume 8, Issue 2, February – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Sample output of the proposed technique is displayed in VI. FUTURE SCOPE
Fig 9.
We’ll also aim to improve the gender classification by
fusing the fingerprint features obtained by the convolution
layers with the minuscule points utilized in the fully connected
layer. We’ll be contrasting our novel approach with other
classifiers in future study. We will look at whether additional
factors, such as fingerprint thickness or valley thickness, may
improve the capability of deep CNN models for classification.

REFERENCES

[1]. Vinod C Nayak et al. “Sex differences from fingerprint

ridge density in Chinese and Malaysian population”. In:
Forensic science international 197.1-3 (2010), pp. 67–
69.
[2]. MD Nithin et al. “Gender differentiation by finger ridge
count among South Indian population”. In: Journal of
forensic and legal medicine 18.2 (2011), pp. 79–81.
[3]. Sajid Ali Khan et al. “A comparative analysis of gender
classification techniques”. In: International Journal of
Bio-Science and Bio-Technology 5.4 (2013), pp. 223–
243.
[4]. Sri Suwarno and P Insap Santosa. “A Short Review of
Gender Classification based on Fingerprint using
Wavelet Transform”. In: International Journal of
Advanced Computer Science and Applications 8.11
(2017).
[5]. Samta Gupta and A Prabhakar Rao. “Fingerprint based
gender classification using discrete wavelet transform &
artificial neural network”. In: International Journal of
Computer Science and mobile computing 3.4 (2014), pp.
1289–1296.
[6]. Ahmed M Badawi et al. “Fingerprint-Based Gender
Classification.” In: IPCV 6.8 (2006), p. l.
[7]. SS Gornale, M Basavanna, and R Kruti. “Fingerprint
based gender classification using local binary pattern”.
In: International Journal of Computational Intelligence
Research, ISSN (2017), pp. 0973–1873.
[8]. Chandrakant P Divate and Syed Zakir Ali. “Study of
different bio-metric based gender classification
systems”. In: 2018 International Conference on Inventive
Research in Computing Applications (ICIRCA). IEEE.
Fig. 9. Sample Output of the Proposed Technique 2018, pp. 347–353.
[9]. Emanuela Marasco, Luca Lugini, and Bojan Cukic.
V. CONCLUSION “Exploiting quality and texture features to estimate age
and gender from fingerprints”. In: Biometric and
Due to the huge potential of fingerprints as an effective surveillance technology for human and activity
means of identification, an attempt has been made in the identification XI. Vol. 9075. SPIE. 2014, pp. 112–121.
current study to analyze the link between fingerprints and [10]. Ashish Mishra and Preeti Maheshwary. “A novel
gender of an individual. The results of gender classification technique for fingerprint classification based on naive
using these dominant traits suggested that this technique might bayes classifier and support vector machine”. In:
be a good contender for use in forensic anthropology to narrow International Journal of Computer Applications 975
the list of prospective suspects and offer a likely probability (2017), p. 8887.
value for a suspect’s gender. [11]. Jitendra Sheetlani, Rajmohan Pardeshi, et al.
“Fingerprint based automatic human gender
identification”. In: Int. J. Comput. Appl 170.7 (2017), pp.
1–4.

IJISRT23FEB1204 www.ijisrt.com 1489

Volume 8, Issue 2, February – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[12]. AS Falohun, OD Fenwa, and FA Ajala. “A
fingerprintbased age and gender detector system using
fingerprint pattern analysis”. In: International Journal of
Computer Applications 136.4 (2016), pp. 0975–8887.
[13]. Mangesh K Shinde and SA Annadate. “Analysis of
fingerprint image for gender classification or
identification: Using wavelet transform and singular
value decomposition”. In: 2015 international conference
on computing communication control and automation.
IEEE. 2015, pp. 650–654.
[14]. V Rekha et al. “Dactyloscopy based gender classification
using machine learning”. In: 2019 IEEE International
Conference on System, Computation, Automation and
Networking (ICSCAN). IEEE. 2019, pp. 1–5.
[15]. Suman Sahu, A Prabhakar Rao, and Saurabh Tarun
Mishra. “Fingerprints based gender classification using
adaptive neuro fuzzy inference system”. In: 2015
International Conference on Communications and Signal
Processing (ICCSP). IEEE. 2015, pp. 1218–1222.
[16]. Philipp Terh¨orst et al. “Deep and multi-algorithmic
gender classification of single fingerprint minutiae”. In:
2018 21st International Conference on Information
Fusion (FUSION). IEEE. 2018, pp. 2113–2120.
[17]. KS Arun and KS Sarath. “A machine learning approach
for fingerprint based gender identification”. In: 2011
IEEE Recent Advances in Intelligent Computational
Systems. IEEE. 2011, pp. 163–167.
[18]. Ronny Merkel, Jana Dittmann, and Claus Vielhauer. “A
first public research collection of high-resolution latent
fingerprint time series for short-and long-term print age
estimation”. In: IEEE transactions on information
forensics and security 12.10 (2017), pp. 2276–2291.
[19]. Shivanand Gornale, Abhijit Patil, and C Veersheety.
“Fingerprint based gender identification using discrete
wavelet transform and gabor filters”. In: International
Journal of Computer Applications 975 (2016), p. 8887.
[20]. Yahaya Isah Shehu et al. “Sokoto coventry fingerprint
dataset”. In: arXiv preprint arXiv:1807.10609 (2018).

IJISRT23FEB1204 www.ijisrt.com 1490