Symmetry Alignment–Feature Interaction Network for Human Ear Similarity Detection and Authentication DOI Open Access

Li Yuan,

H. Zhou,

Jiangyun Li

et al.

Symmetry, Journal Year: 2025, Volume and Issue: 17(5), P. 654 - 654

Published: April 26, 2025

In the context of ear-based biometric identity authentication, symmetry between left and right ears emerges as a pivotal factor, particularly when registration involves one ear authentication utilizes its contralateral counterpart. The extent to which bilateral supports consistent verification warrants significant investigation. This study addresses this challenge by proposing novel framework, Symmetry Alignment–Feature Interaction Network, designed enhance robustness. proposed network incorporates Alignment Module, leveraging differentiable geometric alignment dual-attention mechanism achieve precise feature correspondence ears, thereby mitigating robustness deficiencies conventional methods under pose variations. Additionally, Feature Network is introduced amplify nonlinear interdependencies binaural features, employing difference–product dual-path architecture discriminability through Dual-Path Similarity Fusion. Experimental validation on dataset from University Science Technology Beijing demonstrates that method achieves similarity detection accuracy 99.03% (a 9.11% improvement over baseline ResNet18) an F1 score 0.9252 in tasks. Ablation experiments further confirm efficacy reducing false positive rate 3.05%, combination with shrinking standard deviation distributions negative samples 67%. A multi-task loss function, governed dynamic weighting mechanism, effectively balances learning objectives. work establishes new paradigm for features symmetry, integrating modeling Fusion advance precision authentication.

Language: Английский

Symmetry Alignment–Feature Interaction Network for Human Ear Similarity Detection and Authentication DOI Open Access

Li Yuan,

H. Zhou,

Jiangyun Li

et al.

Symmetry, Journal Year: 2025, Volume and Issue: 17(5), P. 654 - 654

Published: April 26, 2025

In the context of ear-based biometric identity authentication, symmetry between left and right ears emerges as a pivotal factor, particularly when registration involves one ear authentication utilizes its contralateral counterpart. The extent to which bilateral supports consistent verification warrants significant investigation. This study addresses this challenge by proposing novel framework, Symmetry Alignment–Feature Interaction Network, designed enhance robustness. proposed network incorporates Alignment Module, leveraging differentiable geometric alignment dual-attention mechanism achieve precise feature correspondence ears, thereby mitigating robustness deficiencies conventional methods under pose variations. Additionally, Feature Network is introduced amplify nonlinear interdependencies binaural features, employing difference–product dual-path architecture discriminability through Dual-Path Similarity Fusion. Experimental validation on dataset from University Science Technology Beijing demonstrates that method achieves similarity detection accuracy 99.03% (a 9.11% improvement over baseline ResNet18) an F1 score 0.9252 in tasks. Ablation experiments further confirm efficacy reducing false positive rate 3.05%, combination with shrinking standard deviation distributions negative samples 67%. A multi-task loss function, governed dynamic weighting mechanism, effectively balances learning objectives. work establishes new paradigm for features symmetry, integrating modeling Fusion advance precision authentication.

Language: Английский

Citations

0