Bivariate empirical mode decomposition for ECG-based biometric identification with emotional data.

Area of Science:

• Biometrics
• Signal Processing
• Affective Computing

Background:

• Electrocardiogram (ECG) signals are influenced by emotional states, potentially impacting biometric identification accuracy.
• Developing emotion-independent feature extraction methods is crucial for reliable real-world ECG-based biometrics.

Purpose of the Study:

• To evaluate bivariate empirical mode decomposition (BEMD) for extracting emotion-independent features from ECG signals for biometric identification.
• To assess the performance of BEMD-based features under varying emotional conditions.

Main Methods:

• Utilized ECG signals from the Mahnob-HCI database.
• Applied bivariate empirical mode decomposition (BEMD) to ECG signals.
• Extracted features based on statistical distributions of dominant frequencies post-BEMD analysis.
• Employed k-Nearest Neighbors (kNN) classifier with 10-fold cross-validation.

Main Results:

• Achieved 99.5% accuracy in identifying 26 subjects when emotional states were ignored.
• Maintained high accuracy of approximately 99.4% when emotional states were considered.
• Demonstrated high consistency across validation folds.

Conclusions:

• The proposed BEMD-based feature extraction method offers robust, emotion-independent features for ECG biometrics.
• The method shows significant promise for reliable identification in real-world scenarios.
• Further validation with diverse classifiers and ECG signal variations (arrhythmias, age groups, other databases) is recommended.