Table 1 Overview of the relevant works.
From: Securing Internet-of-Medical-Things networks using cancellable ECG recognition
Reference | Number of Subjects | Acquisition method or database | Classifier Type | Performance Metrics | Limitations |
|---|---|---|---|---|---|
Lee and Kwak17 | 1- 100 2- 290 | 1- CU-ECG DB 2- PTB-ECG DB | EECGNet-based SVM | Accuracy = 98.25% | Utilization of merely two datasets. Making use of the initial ECG templates. Additional complexity by transforming ECG signals into images. |
Barros et al.19 | 1- 1500 2- 100 | 2018 database for PhysioNet Computing in Cardiology | RF Classifier | 1-Accuracy = 92% 2-Accuracy = 95% | Just one ECG dataset is used. Making use of the initial ECG templates. Taking noise in the ECG signals into account. |
Su et al.21 | NaN | VeinECG derived from the ECG-ID and FVPolyU finger vein datasets | Discriminant Correlation Analysis (DCA) | Accuracy = 94% | Utilization of just one ECG dataset. Making use of the initial ECG templates. Taking noise in ECG signals into account. |
Zhang et al.22 | 85 | 3 public ECG databases | Matching method | Accuracy = 97.6% | Taking noise in ECG signals into account. Making use of the initial ECG templates. Taking 4 s for authentication and 20 s for registration of a new subject. |
Hammad et al.23 | 1- 25 men, and 22 women signals 2- 290 3- 65 subjects (49 males and 16 females) | 1- MIT-BIH arrhythmia dataset 2- PTB dataset 3- CYBHi dataset | Feed Forward Neural Network (FFNN) | 1-EER = 0.06 2-EER = 0.14 3-EER = 0.09 | Disregarding noise in ECG signals. |
Kim et al.24 | 89 | ECG-ID database | Euclidean detection | Accuracy = 94.3% | Utilization of just one ECG dataset. Taking noise in ECG signals into account. |
Zhao et al.25 | 50 | Database for Physionet ECG | Convolutional Neural Network (CNN) | Accuracy = 99% | Utilization of just one dataset. Making use of the initial ECG templates. Taking noise in ECG signals into account. Additional complexity due to transforming the ECG signal into an image. |
Blasco et al.26 | 25 | Low-cost sensor dataset | One-class classifier with density estimation | Accuracy = 99% EER = 0.16 | Utilization of just one dataset. Making use of the initial ECG templates. Taking noise in ECG signals into account. |
Bugdol et al.27 | 30 | Voice-ECG database | K-Nearest Neighbors (KNN) classifier | Accuracy = 89% | Utilization of just one ECG dataset. Making use of the initial ECG templates. Taking noise in ECG signals into account. |
