Interpretable machine learning for cardiogram-based biometrics
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918331687108608 |
|---|---|
| author | Tanasković, Ilija Lazarević, Ljiljana B. Knežević, Goran Milosavljević, Nikola Dubljević, Olga Bjegojević, Bojana Miljković, Nadica |
| author_facet | Tanasković, Ilija Lazarević, Ljiljana B. Knežević, Goran Milosavljević, Nikola Dubljević, Olga Bjegojević, Bojana Miljković, Nadica |
| contents | This study investigates the role of electrocardiogram (ECG) and impedance cardiogram (ICG) features in biometric identification, emphasizing their discriminative capacity and robustness to emotional variability. A total of 29 features spanning four domains (temporal, amplitude, slope, and morphological) are evaluated using random forest (RF) models combined with multiple interpretability methods. Feature importance shows that both ECG- and ICG-derived features are consistently ranked among the top 10 by Gini importance, permutation importance, and SHAP values, with ECG features, particularly QRS-centric descriptors, occupying the highest positions. In parallel, ICG BCX features contribute complementary, however, with lower cross-method stability. Correlation analysis reveals substantial multicollinearity, where the RF distributes and diminishes importance across highly correlated pairs, confirming reduced independent contributions. Statistical analysis identifies 14 features with significant differences between baseline and anger, without a clear pattern by domain. Feature selection with recursive feature elimination and genetic algorithms converges on a subset (12 features) that attains accuracy within 1% of the full set (99%), improving efficiency in storage and computation. Proposed complementary analyses indicate that the individuality is primarily encoded in the QRS-related ECG features across all four domains. Meanwhile, BCX-derived ICG features contribute mainly through amplitude, providing supportive but less stable discriminatory cues. The confirmed resilience of QRS-centric descriptors to emotional variation, where stable inter-individual differences in the QRS complex could be traced to variations in ventricular mass, conduction pathways, and thoracic geometry, may indicate their central role in future models of cardiogram-based identity. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_19775 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Interpretable machine learning for cardiogram-based biometrics Tanasković, Ilija Lazarević, Ljiljana B. Knežević, Goran Milosavljević, Nikola Dubljević, Olga Bjegojević, Bojana Miljković, Nadica Signal Processing This study investigates the role of electrocardiogram (ECG) and impedance cardiogram (ICG) features in biometric identification, emphasizing their discriminative capacity and robustness to emotional variability. A total of 29 features spanning four domains (temporal, amplitude, slope, and morphological) are evaluated using random forest (RF) models combined with multiple interpretability methods. Feature importance shows that both ECG- and ICG-derived features are consistently ranked among the top 10 by Gini importance, permutation importance, and SHAP values, with ECG features, particularly QRS-centric descriptors, occupying the highest positions. In parallel, ICG BCX features contribute complementary, however, with lower cross-method stability. Correlation analysis reveals substantial multicollinearity, where the RF distributes and diminishes importance across highly correlated pairs, confirming reduced independent contributions. Statistical analysis identifies 14 features with significant differences between baseline and anger, without a clear pattern by domain. Feature selection with recursive feature elimination and genetic algorithms converges on a subset (12 features) that attains accuracy within 1% of the full set (99%), improving efficiency in storage and computation. Proposed complementary analyses indicate that the individuality is primarily encoded in the QRS-related ECG features across all four domains. Meanwhile, BCX-derived ICG features contribute mainly through amplitude, providing supportive but less stable discriminatory cues. The confirmed resilience of QRS-centric descriptors to emotional variation, where stable inter-individual differences in the QRS complex could be traced to variations in ventricular mass, conduction pathways, and thoracic geometry, may indicate their central role in future models of cardiogram-based identity. |
| title | Interpretable machine learning for cardiogram-based biometrics |
| topic | Signal Processing |
| url | https://arxiv.org/abs/2510.19775 |