EdgeEar: Efficient and Accurate Ear Recognition for Edge Devices

2025-02-11

AI Summary

To address the challenge of deploying accurate yet efficient ear recognition models on resource-constrained edge devices, this paper proposes a lightweight hybrid CNN-Transformer architecture. We introduce, for the first time, low-rank approximation in linear layers, reducing model parameters by 50× (<2M). Combined with structured pruning, post-training quantization, and edge-optimized inference techniques, our approach achieves state-of-the-art equal error rate (EER) on the UERC2023 benchmark—matching server-side SOTA accuracy—while significantly lowering computational cost. This work constitutes the first empirical validation of high-accuracy ear biometric recognition on edge hardware, demonstrating practical feasibility for privacy-sensitive, on-device authentication. It establishes a novel paradigm for lightweight, privacy-preserving identity verification at the edge.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Biometrics, Face, Gesture & Pose
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• Security and Privacy: Privacy-enhancing technologies
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Ear recognition is a contactless and unobtrusive biometric technique with applications across various domains. However, deploying high-performing ear recognition models on resource-constrained devices is challenging, limiting their applicability and widespread adoption. This paper introduces EdgeEar, a lightweight model based on a proposed hybrid CNN-transformer architecture to solve this problem. By incorporating low-rank approximations into specific linear layers, EdgeEar reduces its parameter count by a factor of 50 compared to the current state-of-the-art, bringing it below two million while maintaining competitive accuracy. Evaluation on the Unconstrained Ear Recognition Challenge (UERC2023) benchmark shows that EdgeEar achieves the lowest EER while significantly reducing computational costs. These findings demonstrate the feasibility of efficient and accurate ear recognition, which we believe will contribute to the wider adoption of ear biometrics.

Problem

• Deploy ear recognition on resource-limited devices
• Reduce model size while maintaining accuracy
• Achieve high efficiency and low error rates

Innovation

• Hybrid CNN-transformer architecture
• Low-rank linear layer approximations
• Sub-two million parameter model