Fake Currency Detection using Image Processing

Srishti Sharma, Shreyansika Jain, Saurabh, Satyam Kuhar

Department of Computer Science and Engineering
The NorthCap University Gurugram, India
[email protected], [email protected],
[email protected], [email protected]

Abstract—The continuous advancements in technology make The volume of daily transactions in commercial places like
the process of production and dissemination of counterfeit banks, malls, jewelry stores, etc. is very high. Such locations
currency easier than ever before. Counterfeit currency may be generally purchase machines that apply UV light and other
produced with the intent of making profit, engaging in illegal methods to check the currency's legitimacy. However, it can
activities such as drug trafficking, money laundering or
be quite difficult for the common man to simply tell whether
financing terrorism, all of which hurt society greatly. Most of
the time, innocent individuals become the medium for a piece of currency is real or not, by looking at it or handling
circulation of counterfeit currency who inadvertently circulate it and they may suffer losses, particularly during bank
2024 International Conference on Intelligent Systems for Cybersecurity (ISCS) | 979-8-3503-7523-7/24/$31.00 ©2024 IEEE | DOI: 10.1109/ISCS61804.2024.10581207

counterfeit currency without realizing it is fake, especially if deposits or other transactions. This research work tries to
they are not familiar with the security features of legitimate address this gap by proposing a low-cost, simple software
currency. Banks and other financial or commercial solution that can be used by anyone, allowing them to check
organizations have equipment in place to verify the legitimacy the currency they have for authenticity. The approach
of currency. However, the common man does not have access to described in this work can be converted into a simple app to
such systems. Hence there is an urgent and dire need for a make it available to the masses. Additionally, this system can
software that can be utilized by the common man at any place
be extended to be able to work on a number of currencies
and time to identify counterfeit currency. This paper proposes
an approach employing the image processing technique, across the globe. The methodology is built around image
Oriented Fast and Rotated Brief (ORB) to help the common processing, in which a variety of stages are performed to
man identify counterfeit currency. The approach shows process a currency's picture and determine if it is authentic or
promising results achieving up to 0.7674 SSIM score. not. In this approach will allow a user to check multiple
currencies in one place, eliminating the need for multiple
Keywords— Counterfeit currency, Image Processing, Python software programmes for different currencies. Our machine
programming language, Machine Learning, Deep Learning learning will essentially have a high accuracy rate because it
will check every aspect that assists a normal person in
I. INTRODUCTION detecting different currencies that can either be fake or real.
If the result is positive, the output will be "the currency is
real," and if it is negative, the output will be "the currency is
For the financial exchange of goods, almost all nations fake."
employ different currencies. With the incessant
advancements in technology, counterfeit currency is II. LITERATURE REVIEW
produced using modern printers and computer editing
software. Currency can be generated and replicated easier We carry out a detailed literature survey of the field of fake
than ever before, and it is extremely difficult to differentiate currency detection. Authors in [3] outline a technique to
between fake and legitimate currency. Fake currencies are enhance the currency checking majorly in money-oriented
simply slipping away into bundles of real money. The areas like Multinational Corporations (MNCs) etc. Multiple
presence of counterfeit money affects nations in many ways pre-processing ideas were introduced such as radiometric
like sponsoring illegal activities such as drug trafficking, correction and geometric correction to rectify spectral errors
money laundering or financing terrorism. To combat the or contortion due to analyzing earth geometric dissimilarity
menace of fake currency, governments, and central banks etc. Different articles were distinguished, and conclusions
across the globe, implement various security features in their were drawn based on the accuracy rates reached utilizing
currency, educate the public about these features, and enforce different methods that were provided.
laws against counterfeiting. In [4], the authors have suggested that based on the various
features that were extracted for the comparison, a method for
Despite all these measures, there are many instances of identifying the counterfeit cash was discovered. Different
circulation and production of counterfeit currency. For approaches, like the usage of feature vectors to store the
instance, in Agra, GRP busted a counterfeit currency printing extracted features, etc., were reached at various stages of
network [1]. On December 10, 2022, the Kota-Patna Express histogram equalization. The security thread, RBI micro-print,
was busted by the railway police for carrying 1.5 lakh and serial number were the features that were extracted for
counterfeit 500- rupee bills and on [2] December 8, 2022, two use in detection.
people in their 30s or 40s used counterfeit £50 Scottish notes In [5], by contrasting visual elements like the distance
to purchase a variety of products, including clothing, jewelry, between visuals of Mahatma Gandhi and other notations, the
and shoes. authors have given a way to check that the currency is fake or
not. It will be helpful for a system that only uses software
processing.

Authorized licensed use limited to: Zhejiang University. Downloaded on November 15,2024 at 05:56:03 UTC from IEEE Xplore. Restrictions apply.
 In [6] Indian currency counterfeiting is illustrated by Journal has suggested that based on the various features that
Amandeep Kaur. Three key elements in the research article were extracted for the comparison, Adaptive HARRIS
below were: the inactive image, the RBI logo, and the value Corner Detection by applying feature corner vectors to store
numeral and rupee symbol in color. These three key features many extracted features.
were utilized and applied to an algorithm that recognizes In [14], Nadia al Basttaqi and Adel Atiq published paper
counterfeit money. named Dirham verification system relies on image processing
Thigale et al. [7] proposed that a user take and note the to extract characteristics. Analogue image processing used
counterfeit currency, which will then be detected using for the identification using currency. Image performance
image processing. They will use MATLAB algorithms to measurements were used as a fraud detection approach
detect the value of a note, which will then be implemented In [15], the author Peter Angelou and Ben Laughlin in US
on the counterfeit the detecting machine contains a UV LED Dollar Recognition System review paper in Las Vegas
and a photodiode. Journal has suggested that based on the various features that
Vipin Kumar Jain [8] has examined digital image processing were extracted for the comparison, Adaptive HARRIS
with the help of android cellphone to check the fake currency. Corner Detection by applying feature corner vector to store
The photo taken by the android phone was saved in the phone many extracted features.
and converted to grayscale design, binaries image and some Dr. George Camus and Jules Hugo [16] carried out Swizz
morphological operations were developed on this application. Franc Identification. They had used ILSVRC 2011 currency
They then had taken the block of each substance for dataset. Dataset which has 5000 images of Swizz Franc and
completing the image using various application of hand- then implemented modified YOLO + modified AlexNet
written character feature extraction. techniques.
Prathap [9] have used various image and applying Author in [17] worked on fake currency to detect the
preprocessing techniques. They have used the image crop, watermarking, physically variable ink, protection thread,
smooth and adjusted for preprocessing the data of various opaque image and approaches like counterfeit monitoring
images. Then converted the image to gray scale and applied pens identified using machine learning models. Multiple
image segmentation after that. finally compared the image to methods were used to identify the money. The phases
original and the fake one. consist of edge detection, extraction of features, visualize
Gajanan et al. [10] in this research paper outlines many segmentation, grayscale transformation, and image
methods that were previously used to identify counterfeit comparison.
money and explain why those technologies were not available In [18], The authors developed a Hyperspectral Imaging
to everyone. The main focus of the section was on the usage (HSI) technique utilizing the Raspberry Pi platform. They
of counterfeit money and a new method will be introduced, used the Mean Grey Values (MGV) to construct a model
and since it will be a mobile application, it might be used that compares the currency's major attributes to counterfeits.
frequently to spot counterfeit money. The method we can use The outcome is based upon the lower wavelength value of
to build a system to detect counterfeit money, particularly MGV if it is between 400 nm to 500 nm so the currency is
Indian currency. The process involved taking a picture of the not counterfeit.
money, turning it to grayscale, and then using a machine In [19], Shamini et al. conducted a study where they focused
learning model that has been trained to detect false currency. on color detection, edge detection, and segmentation of
Here the key features to detect if it's fake or not are Security currency. Their goal was to extract features and compare
Thread, Serial Number, Latent Mark. Then they have them with counterfeit currencies in order to obtain results.
maintained an intensity threshold of 70%; if it has exceeded They utilized the SSIM score as a metric to evaluate their
the currency will not be counterfeit. findings.
Aakash Patil [11] came up with a fresh strategy to enhance The researchers that have studied the identification of fake
the speed and accuracy of detecting Indian currency currency have used a variety of models and strategies for
transactions. It consists of the OpenCV libraries, real-time effective detection, but several gaps and opportunities have
computer vision, which is used for tasks like segmentation not been fully explored. The main purpose of this endeavour
and currency recognition, as well as the NumPy module for was to develop a simple, cost-effective and easy-to-use
numerical processing. solution for the common man. Compared to previous
Authors in [12] worked on detection of fake Libyan currency, suggested techniques, this system takes into consideration
and the procedure suggested by them begins with data more features for comparisons, making it more robust.
collection using all of the original currency and this data is Additionally, it highlights currency differences, rather than
subsequently utilized to train the machine learning algorithm. just displaying the outcome.
In this model, a picture of money was first taken, and during
the pre-processing stage, it was enhanced—a process known Almost all the research in the direction of counterfeit
as "image restoration." The Red, Green and Blue (RGB) is currency detection focuses only on Indian currency. As a
then turned into grayscale. Then used the Sobel Operator further extension to the current work, we are now
Algorithm to do image segmentation. The new picture implementing it on other currencies like dollars, euros, etc. to
features and the features of the real money that had already make the proposed technique globally applicable. The
been calculated were compared during the feature extraction approach proposed could be used to create an application
stage. To tell if the currency is real or counterfeit, use Mean which is user friendly and can be installed in mobile phones
Squared Error(MSE) and Root Mean Squared Error(RMSE). making it.
In [13], the author Peter Angelou and Ben Laughlin in US
Dollar Recognition System review paper in Las Vegas

Authorized licensed use limited to: Zhejiang University. Downloaded on November 15,2024 at 05:56:03 UTC from IEEE Xplore. Restrictions apply.
 TABLE I. STRENGTHS AND WEAKNESSES Aakash (2019) CNN classifiers Only use a simple
used to detect fraud machine learning
Authors Strengths Weaknesses currency. model.
P.Julia, A. Using Matlab to Model can only Simple and easy to Not verifying
Sheema store the files and be used on 500 use. feature wise
(2016) edge detection and 1000 detection
algorithm to reduce demonetised Trupti, Dr. Using the neural Number panel and
the amount of data currency. N.G Bawane network and image Mahatma Gandhi
in images. It cannot be (2011) processing to check photo are key
An average accuracy utilized for the features. features which
of 94.72% in the international Using Matlab for were not utilized
whole model. payments. edge detection and while making the
Komal, Ami, Using local spatial Model use feature capture. model.
Jay (2015) features for image demonetized Data while making
recognition. currency the model is too
Use of Two Only verify the low.
dimensional common aspects. Ji Qian, MATLAB to detect Images should be
discrete wavelet Dongping features. in specific types
transform (2D Qian, Mengjie such as BMP,
DWT) for optimal Zhang (2006) HDF, JPEG, PCX,
and efficient model. TIFF, XWB, PNG.
Mayadevi, Using image Not a general
Vaijinath, processing and modal that can be Peter , Ben Adaptive HARRIS Training dataset is
Vaibhav binarization. used for different (2006) Corner Detection by very small which
(2017) Model is used on currency notes applying feature leads to less
the latest currency. (like 10,20 corner vectors to accuracy.
50,100,200 rs store many
notes) can only be extracted features.
used for 500 and Dr. George, Used modified Use of basic
2000 notes only. Jules Hugo YOLO + modified models.
Amandeep Using MATLAB Trained the model (2015) AlexNet techniques.
(2019) RGB channels for only using 30 input Prasad et al. Cost efficient and It cannot be
calculating three images and having (2023) less time consuming utilized for
different features an accuracy of by using machine international
only 23 images. learning model. payments and is
Prof. Madhav, Using algorithms Uses a basic not supported on
Lina (2017) Structure Similarity machine learning every device.
Index Metric model. Mukundan Et Using a Raspberry The data for
(SSIM), Gray Level Model use al. (2023) Pi camera to detect comparison is too
Co-occurrence demonetized the ink marks on the low so the
Matrix (GLCM) currency currency. outcome might
Use ANN to detect have some false
currency amount true values in
Vipin Jain Using SSIM and a The data used to future.
(2019) neural network build the model is Shamini et al. Using edge Use of basic
model to check the low and can lead to (2024) detection and models.
features. some false results segmentation to
when notes photos check the features.
are not given in
High-definition
quality. III. PROPOSED WORK
Deborah Performance It uses a manual
Soniya (2014) metrics used Peak model rather than
Signal-to-Noise an automatic A. Dataset Description:
Ratio (PSNR),Mean model. The dataset used for currency detection comprises Indian
Squared Error Rate
(MSE),Root Mean Currency notes of denomination 500 and 2000. The main aim
Squared Error Rate of the dataset is to have the images of different currency notes
(MSE) both fake and real. This dataset is used to train the model
Tushar, For feature Need a large consisting of a repository and is custom built by collecting
Gajanan, extraction it uses storage system to
Pratik and P MATLAB. run the MATLAB various features of currency.
Chitra (2017) Simplicity and high model. The dataset contains the following repositories:
performance speed. Slow detection Sub dataset for Rs. 500 currency notes: Image of real notes,
process if not Image of fake notes, Multiple images of each security feature.
suitable processor
found. Sub dataset for Rs. 2000 currency notes: Image of real notes,
Abdulkareem, Using edge Using the Image of fake notes, Multiple images of each security feature.
Ahmed (2019) detection to check most basic The several security features that are being considered are:
the features. approach, how to Rs.500 in Devanagari and English script (2 features), Ashoka
RMSE to get the check the
result. currency.
pillar Emblem (1 feature), RBI symbols in Hindi and English
Did not check (2 features),500 rupees written in Hindi (1 feature), RBI logo
every essential (1 feature), Bleed Lines on Left and Right side (2 features),
feature of the Number Panel (1 feature). The same are outlined in Figures
currency.
1, 2 and 3.

Authorized licensed use limited to: Zhejiang University. Downloaded on November 15,2024 at 05:56:03 UTC from IEEE Xplore. Restrictions apply.
 For feature 8-9:
This is the explicit feature which is only present in the Indian
currency the region where the bleed line are present. There
are two bleeding line right and left edge of the currency. For
Rs. 500 currency there are 5 lines and for Rs. 2000 there are
7 lines present on both the side. First, the algorithm for the
given feature counts and verify the number of bleed line
present on each side by snipping the input image. Then
further image thresholding is done using the suitable value.
Fig 1. Features of Rs.500
The thresholding verifies that only black bleed line present
on the white background. Then we calculate the number of
bleed lines present should be same as the currency type.

For feature 10 (For all the currency):

This feature is present in every currency there is number
panel in the bottom right part where the serial number of
every currency note is present. There should be total 9-
character present on the number panel. First, thresholding is
done so that all the black character can only be remaining on
the white background and easy to mark. Further we perform
Fig 2. Features of Rs.2000
contour detection for the number panel detected by the
thresholding. Third, we do boundary rectangle around each
contour. The erroneous are also eliminated that have too
much noise present in the image. Lastly, all the remaining
rectangles are calculated that gives us the proper character
detection of the thresholding image.

C. Implementation:

SSIM (Structural Similarity Index):

SSIM (Structural Similarity Index) [20] is a metric used to
Fig 3. Features of US Dollar measure the similarity between two images. It is commonly
used in image processing and computer vision applications,
B. Design Methodology: where it is used to evaluate the quality of compressed or
processed images. In this article, we will describe the SSIM
Pre-processing: metric in detail, including its components, working principle,
The preprocessing is done of the input image. In this and applications.
preprocessing first the input image is resized to the fixed size Components of the SSIM Metric: The SSIM metric consists
(1167px, 519px). Fixed size help to make the computations of three key components: luminance, contrast, and structure.
easier. Image smoothening is done with the help of Gaussian Luminance refers to the brightness of an image and is
Blurring Method. Gaussian Blurring Method help to remove calculated as the average pixel intensity. The luminance
noise and increases the capability of the system. component of the SSIM metric compares the average pixel
Gray Scale conversion is done because an RGB image intensities of two images. Contrast refers to the difference in
contains 3 channels but the gray scaled image uses only one pixel intensity between adjacent pixels in an image. The
channel. By doing this it helps to make the processing and contrast component of the SSIM metric compares the contrast
computation easy rather than in the processing normal image. of the two images. Structure refers to the spatial relationship
between pixels in an image. The structure component of the
For feature 1-7 or feature 1-2 (of US Dollar): SSIM metric compares the structure of the two images.
After performing the preprocessing of the input image and Working Principle of the SSIM Metric: The SSIM metric
converting into the gray scale feature detection is done. In this compares two photos structural similarities through the
feature detection all the main features are detected and are luminance, contrast, and composition of the photos. The
further matched using Oriented FAST and Rotated BRIEF SSIM metric is calculated as a combination of three
(ORB). Then further features extraction is done with the help components: luminance, contrast, and structure.
of the ORB location for each template that has been detected Sensitivity to Structural Changes: Traditional metrics like
of the given input image. The features are highlighted with a MSE and PSNR are highly sensitive to any differences
box around the features. These highlighted areas are further between images, including minor variations in pixel values.
cropped by craving into a 3D pixel matrix of the image. We SSIM, on the other hand, focuses on capturing significant
will apply gray scaling and Gaussian blur to smoothen the changes in structure, making it more robust to small
image and now the feature is prepared to be compared to the variations and noise.
corresponding feature in the trained model. Further we will
be assigning the SSIM score to each feature

Authorized licensed use limited to: Zhejiang University. Downloaded on November 15,2024 at 05:56:03 UTC from IEEE Xplore. Restrictions apply.
ORB:
A popular feature detection and extraction approach in
computer vision and image processing called ORB (Oriented
FAST and Rotated BRIEF) is utilized for tasks including
object recognition, tracking, and 3D reconstruction. It was
first presented by Ethan Ruble in the year 2011.
A corner recognition system called FAST (Features from
Accelerated Segment Test) finds important points in a picture
based on their intensity and gradient. The FAST algorithm
can quickly identify key points in real-time applications and
is created to be computationally efficient. A feature
descriptor called BRIEF (Binary Robust Independent
Elementary Features) creates a binary code to express the
intensity connections between pairs of pixels in the
immediate vicinity of a key point. The BRIEF method
quickly and effectively creates a condensed representation of
the local visual structure. Working Principle of the ORB
Algorithm. The ORB algorithm works by detecting and
describing features in an image, which can then be used for
various computer vision tasks such as object recognition,
tracking, and 3D reconstruction. The ORB algorithm uses a
scale pyramid to detect key points at different scales and
orientations, and then applies a modified version of the
BRIEF descriptor to compute a unique binary code for each
key point.

Fig 5. Flow chart for US Dollar

IV. RESULTS

Being able to use money is essential to surviving, thus

maintaining its originality is always important. Since paper
money is used much more frequently, a technique to identify
counterfeit money is required. As the new currencies are
employed in commerce, the suggested system appears to be
helpful in determining if the currency is legitimate or not. The
model approach which was used will essentially have a high
accuracy rate.It will check every aspect that assists a normal
person in detecting different currencies that can either be fake
or real. The model will examine every characteristic of the
currency, such as the size of its digits or any special patterns
that are invisible to the human eye, and compare it to the
original data to determine whether the input is genuine or not.
This will be a better choice because there are several apps for
various currencies, but maintaining their use will be easier
with an integrated system that is highly accurate.
TABLE II. FEATURES (1-7) TABLE FOR INDIAN CURRENCY

Features Average SSIM Score Maximum SSIM Score

Feature 1 0.667 1.000
Feature 2 0.667 1.000
Fig 4. Flow chart for Indian Currency Feature 3 0.683 1.000
Feature 4 0.510 1.000
Feature 5 0.670 1.000
Feature 6 0.667 1.000

Authorized licensed use limited to: Zhejiang University. Downloaded on November 15,2024 at 05:56:03 UTC from IEEE Xplore. Restrictions apply.
 Feature 7 0.510 1.000 [2] “Two men spent fake £ 50 Scottish notes in series of shops in a Welsh
town centre”, WalesOnline, 8 December2022,at11:19,
TABLE III. FEATURES (8,9) TABLE FOR INDIAN CURRENCY https://www.walesonline.co.uk/news/wales-news/two-men-spent-
fake-50-25702227
Features Number of lines Maximum Number of
Present lines
[3] P. Julia Grace, Ph.D., A. Sheema, ―A survey on Fake Indian Paper
Feature 8 5 5
Currency Identification System Grace et al., International Journal of
Feature 9 5 5 Advanced Research in Computer Science and Software Engineering
6(7), July- 2016, pp. 340-345 ISSN: 2277 128X
TABLE IV. FEATURES TABLE FOR FOREIGN CURRENCY
[4] Komal Vora, Ami Shah, Jay Mehta, ―A Review Paper on Currency
Features Average SSIM Score Maximum SSIM Score Recognition System‖ International Journal of Computer Applications
Feature 1 0.910 1.000 (0975 – 8887), Volume 115 – No. 20, April 2015
[5] Mayadevi A.Gaikwad, Vaijinath V. Bhosle Vaibhav D Patil,
Feature 2 0.910 1.000
Automatic Indian New Fake Currency Detection Technique [2017]
[6] “Recognition of Fake Indian Currency Note Using Image Processing”
TABLE V. FEATURE 10 TABLE FOR BOTH CURRENCY presented by Amandeep Kaur in 2019
[7] Prof. Madhav Thigale, Lina Ladhane “Note to Coin converter using
Features Number of Maximum Number of Digital Image Processing” [2017].
Characters Present Characters
[8] Vipin Kumar Jain “Recognition of Fake currency detection done by
Feature 10 10 10 image processing technique”[2019]
(Indian [9] Deborah. Soniya Prathap “Detection of Fake currency using Image
Currency) Processing” [2014]
Feature 10 10 10 [10] Tushar Agasti, Gajanan Burand, Pratik Wade and P Chitra “Fake
(US Dollar) currency detection using image processing” [2017]
[11] Abdulkream I. Salem a, Ahmed B. Elmadani “Fake Currency Detection
V. CONCLUSION AND FUTURE WORK using Image Processing (Case Study Libyan Currency)” [2019]
[12] “Indian Paper currency detection” presented by Aakash Patil in 2019
[13] Ms. Trupti Pathrabe, Dr. N.G Bawane Feature Extraction Parameters
This paper proposed a mobile, portable solution in the form
for Genuine Paper Currency Recognition & Verification ‖ International
of an application for the identification of counterfeit money
Journal of Advanced Engineering Sciences and Technologies ‖ Vol No.
using image processing. The proposed solution will allow a 2, Issue No. 1, 085–089, 2011.
user to check multiple currencies in one place, eliminating the [14] Ji Qian, Dongping Qian, Mengjie Zhang ―A Digit Recognition
need for multiple software programmes for different System for Paper Currency Identification Based on Virtual
currencies. This would provide a cost-effective and viable Instruments‖ IEEE Transactions, 1-4244-0555-6/06,2006.
solution for the common man or for small merchants, as the [15] US Dollar Recognition System review paper in Las Vegas Journal
flow of counterfeit money begins with them because they Peter Angelou and Ben Laughlin 2006
lack a thorough understanding of each feature of the [16] Swiss Franc Identification in Paris Dr. George Camus and Jules Hugo
currencies and unwittingly fall victim to scams. In the future, 2015
this work can be extended by integrating counterfeit currency [17] Detection of Fake Currency using Machine Learning Technique –
detectors with cash handling systems including ATMs, Bhumika D, Harsha Kumari, Apoorva H S, Rakshitha S Prasad, Dr
Roopa H, Dr Vani V || International Journal of Engineering Research
vending machines, and point-of-sale (POS) systems. By
& Technology (IJERT) ISSN: 2278-0181 ,Vol. 12 Issue 05, May-2023
providing real-time banknote verification during
[18] Automatic Counterfeit Currency Detection Using a Novel Snapshot
transactions, this integration would lower the danger of Hyperspectral Imaging Algorithm – Arvind Mukundan , Yu-Ming
accepting fake money. All of these updates can be made in Tsao , Wen-Min Cheng, Fen-Chi Lin and Hsiang-Chen Wang || 10
the future to make it a more effective and accurate platform February 2023 the Special Issue Hyperspectral Imaging and Sensing.
for all small merchants. [19] Fake Currency Detection by S. Sweetline Shamini, G. Nandhini, R.S.
Varshini and T. Dharini || Volume 491, 2024, International Conference
on Environmental Development Using Computer Science (ICECS 24)
REFERENCES [20] “Structural Similarity Index Measure”, Wikimedia Foundation, last
modified 12 February 2024, at 18.12,
[1] “GRP busts racket in fake currency in UP’s Agra division”, Times of India, 10 https://en.wikipedia.org/wiki/Structural_similarity_index_measure.
December 2022, at 00:46
(IST), https://timesofindia.indiatimes.com/city/lucknow/grp-busts-racket-in-fake-
currency-in-ups-agra-division/articleshow/96116666.cms

Authorized licensed use limited to: Zhejiang University. Downloaded on November 15,2024 at 05:56:03 UTC from IEEE Xplore. Restrictions apply.