struggle to extract robust features, leading to high false positive rates and low accuracy. This research developed

a feature-fusion authentication system for academic library access control using multi-feature extraction

techniques. 324 university students fingerprint dataset from 81 subjects were captured. The acquired dataset was

preprocessed (cropped, contrast adjustment, gray scale, binarization). The Cross Number Algorithm (CNA) and

Principal Component Analysis (PCA) were used for feature extraction. The Weighted Sum Rule was used to fuse

extracted features from CNA and PCA, generating a unified feature vector. Random Forest Classifier was

employed for classification. The results show that CNA–PCA based system achieved accuracy of 96.91%, CNA

achieved accuracy of 94.14% and PCA produced accuracy (92.59%).

Keywords: Fingerprint-based authentication systems, Academic libraries, Cross Number Algorithm, Principal

Component Analysis, Weighted Sum Rule

manage physical and digital resources for efficient, real-time data access, saving time for both librarians and

users by automating processes and providing instant search capabilities. A typical Library management network

diagram, figure 1, showing multiple computer Systems connected through a library network [6]. The increasing

reliance on digital resources in academic libraries has heightened the need for secure and efficient authentication

mechanisms to regulate access. Traditional methods, such as passwords and ID cards, are increasingly vulnerable

to security breaches, including phishing, theft, and unauthorized duplication [34]. However, recently biometrics

have gained the attention of researchers due to their robustness. The authors in [22] and [5] defined biometrics

as the automatic recognition of individuals based on their physiological and/or behavioral characteristics.

Biometrics can be behavioral characteristics (e.g. voice, signature, and keystroke) and/or physiological

biometrics (finger, Iris, retina, hand, and face) [24] [2]. Due to inadequacies of the traditional methods, biometric

Fingerprints (as shown in figure 2) are the patterns formed on the epidermis of the fingertip. Fingerprints are

made up of a series of ridges and valleys (also called furrows) on the surface of the fingertip and have a core

around which patterns like swirls, whorls, loops, or arches are curved to ensure that each print is unique.

Fingerprints are the patterns formed on the epidermis of the fingertip. Fingerprints are made up of a series of

ridges and valleys (also called furrows) on the surface of the fingertip and have a core around which patterns

like swirls, whorls, loops, or arches are curved to ensure that each print is unique [12] [13] [28].

Figure 2: Sample of Fingerprint [12]

Despite its advantages, single algorithm-feature fingerprint authentication systems face several challenges that

can compromise their effectiveness. Variations in fingerprint quality, such as dry or wet skin, scars, or partial

biometric data. By combining features from different algorithms, feature fusion can mitigate the weaknesses of

individual methods, resulting in improved recognition accuracy and security [38]. For instance, combining

texturebased features with minutiae-based features can enhance the system’s ability to handle lowquality

fingerprints. Fusion compensates for these limitations by integrating additional biometrics or cascade or hybrid

feature extraction techniques, resulting in a system with a higher recognition rate and greater resilience to

variations [8]. Fusion in biometric systems, shown in Figure 3, can be categorized into four primary levels,

sensor-level fusion occurs when raw data from multiple sensors or different biometric traits are combined before

feature extraction [20], [25]; feature-level fusion which integrates feature sets from different biometrics or hybrid

feature extraction methods before the matching process [26], score-level fusion in which individual match scores

from different sensors or algorithms are combined into a single score [35]; and decision-level fusion combines

the final decisions (match or no match) from each modality to arrive at a single authentication decision [4], [16].

without compromising the distinctiveness of the fingerprint data [19]. And to combine the extracted features, the

Weighted Sum Rule fusion technique is applied. This method assigns specific weights to each feature based on

its reliability and contribution to the overall recognition process, ensuring a balanced and comprehensive

representation of the fingerprint data [17].

The rest of the paper is organized as follows: section II contains the related works; section III contains the

research materials and method; section IV entails the results and discussion while section V contains the

conclusion.

In recent years, biometric authentication systems have seen significant advancements, with a strong focus on

multi-feature extraction and fusion techniques to improve accuracy, security, and reliability. In 2009, the

i.e face and palmprint. The both face and palmprint feature are represent by feature code, namely FPCode.

FPCode uses fixed length 1/10 bits coding scheme that is very efficient in matching, and at the same time

achieves higher accuracy than other fusion methods available. This approach compares with the Gabor plus PCA

and Gabor plus KDRC. Experimental results show that both feature level and decision level strategies achieve

much performance with the accuracy of 91.52% and 91.63%. in 2012, researchers in [38] discusses a new patron

authentication approach in an automated and modern library based on the biometric recognition. The person

working at the entrance or at the circulation desk of the library needs to either confirm or determine the identity

of an individual requesting service in library. The purpose of such authentication system is to ensure that the

rendered services are accessed only by a valid user, and not any. The authors of [32] in 2019 proposed system

uses face recognition for entering and getting the details of an end user. Through this application, the book gets

In 2020, the researchers in [27] purposed secure library transaction supported on fingerprint recognition. The

academic libraries can make use of the advantage of the fingerprint recognition technology. The registered person

fingerprint and palm print modalities at the feature level. The authors used a hybrid feature extraction technique,

combining Gabor filters and Local Binary Patterns, to extract unique features from both fingerprints and palm

prints. Support Vector Machine algorithm was employed for classification. The system was evaluated on the

FVC2004 fingerprint dataset and the Poly U palm print dataset, achieving an accuracy of 97.3%. While a study

by [23] in 2022 focused on developing a feature-fusion authentication system specifically for library access

control. The system combined three biometric modalities: face, fingerprint and voice. The authors used a deep

learning framework with Convolutional Neural Networks for feature extraction and a decision tree algorithm for

classification. The system was trained on a custom dataset of 200 users, achieving an accuracy of 98.2%.

In 2022, authors in [30] introduced a continuous authentication system that relied on behavioral biometrics, such

as keystroke dynamics and mouse movement patterns. The authors used a machine learning approach, with

2025, proposed a large-scale feature extraction method based on the correlation analysis of two physicochemical

properties of amino acids: hydrophobicity and hydrophilicity, as well as the correlation between amino acids.

The new approach to feature extraction is OTE-24 approach and Random Forest is used for classification. The

results of this research, evaluated using databases commonly utilized in previous studies, show an accuracy

improvement of over 2.58% compared to existing methods.

Materials

This section entails the techniques used for this work which include Principal Component Analysis (PCA), Cross

Number Algorithm (CNA), weighted sum algorithm (WSA) and Random Forest (RF).

the orientation and frequency of patterns within the biometric data. Specifically, the algorithm utilizes the

concept of cross-number features, which are calculated based on the pixel connectivity in the fingerprint. The

Contrast Enhancement: The contrast of fingerprint images was improved by using adaptive histogram

equalization. This technique adjusts the brightness of pixels in small sections of the image, making it easier to

brightness variations critical for distinguishing fine details in the fingerprint pattern.

Morphological Operations: like erosion and dilation were removed. These processes preserved critical ridge

details while removing isolated noise pixels, enhancing the fingerprint structure’s integrity for more accurate

analysis.

Binarization the grayscale images were converted into a binary format. This binarization step transformed each

features. Mathematically, binarization is represented as follows.

(6)

Feature extraction utilizes CNA, which is widely regarded as an effective approach to identifying ridge endings

and bifurcations, or minutiae, in fingerprint images. This algorithm computes the number of transitions from 0

(7)

completes the loop. This calculation yields a crossing number of one for ridge endings and

three for bifurcations, allowing CNA to isolate these features effectively.

(8)

maximization.

Step 3: Eigenvalue and Eigenvector Computation

eigenvectors v