Multimodal biometric authentication based on deep fusion of electrocardiogram (ECG) and finger vein
Biometric identification depends on the statistical analysis of the unique physical and behavioral characteristics of individuals. However, a unimodal biometric system is susceptible to different attacks such as spoof attacks. To overcome these limitations, we propose a multimodal biometric authenti...
Saved in:
Published in | Multimedia systems Vol. 28; no. 4; pp. 1325 - 1337 |
---|---|
Main Authors | , , |
Format | Journal Article |
Language | English |
Published |
Berlin/Heidelberg
Springer Berlin Heidelberg
01.08.2022
Springer Nature B.V |
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Biometric identification depends on the statistical analysis of the unique physical and behavioral characteristics of individuals. However, a unimodal biometric system is susceptible to different attacks such as spoof attacks. To overcome these limitations, we propose a multimodal biometric authentication system based on deep fusion of electrocardiogram (ECG) and finger vein. The proposed system has
three
main components, which are biometric pre-processing, deep feature extraction, and authentication. During the pre-processing, normalization and filtering techniques are adapted for each biometric. In the feature extraction process, the features are extracted using a proposed deep Convolutional Neural Network (CNN) model. Then, the authentication process is performed on the extracted features using
five
well-known machine learning classifiers: Support Vector Machine (SVM), K-Nearest Neighbors (KNNs), Random Forest (RF), Naive Bayes (NB), and Artificial Neural Network (ANN). In addition, to represent the deep features in a low-dimensional feature space and speed up the authentication task, we adopt Multi-Canonical Correlation Analysis (MCCA). We combine the two biometric systems based on ECG and finger vein into a single multimodal biometric system using feature and score fusion. The performance of the proposed system is tested on
two
finger vein (TW finger vein and VeinPolyU finger vein) databases and
two
ECG (MWM-HIT and ECG-ID) databases. Experimental results reveal improvement in terms of authentication performance with Equal Error Rates (EERs) of 0.12% and 1.40% using feature fusion and score fusion, respectively. Furthermore, the authentication with the proposed multimodal system using MCCA feature fusion with a KNN classifier shows an increase of accuracy by an average of 10% compared with those of other machine learning algorithms. Therefore, the proposed biometric system is effective in performing secure authentication and assisting the stakeholders in making accurate authentication of users. |
---|---|
ISSN: | 0942-4962 1432-1882 |
DOI: | 10.1007/s00530-021-00810-9 |