ProFi-Net: Prototype-based Feature Attention with Curriculum Augmentation for WiFi-based Gesture Recognition

2025-04-28

AI Summary

To address low accuracy and poor generalization in Wi-Fi-based few-shot gesture recognition—caused by scarce training data and sparse signal features—this paper proposes a prototype-driven metric learning framework. Methodologically, it introduces (1) a feature-level attention mechanism coupled with a dynamically weighted distance metric to adaptively optimize discriminative distances among prototypes, and (2) a novel curriculum-style progressive Gaussian noise augmentation strategy applied exclusively to the query set, enhancing model robustness and mitigating overfitting. Evaluated across multiple real-world scenarios, the proposed method achieves an average classification accuracy improvement of 5.2–9.7% over standard prototypical networks and state-of-the-art few-shot approaches, while accelerating training convergence by 38%. The framework thus effectively balances accuracy and efficiency in resource-constrained Wi-Fi sensing tasks.

Technology Category

• Machine Learning: Learning with Manifolds
• Search and Optimization: Learning to Search
• Computer Vision: Motion & Tracking

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This paper presents ProFi-Net, a novel few-shot learning framework for WiFi-based gesture recognition that overcomes the challenges of limited training data and sparse feature representations. ProFi-Net employs a prototype-based metric learning architecture enhanced with a feature-level attention mechanism, which dynamically refines the Euclidean distance by emphasizing the most discriminative feature dimensions. Additionally, our approach introduces a curriculum-inspired data augmentation strategy exclusively on the query set. By progressively incorporating Gaussian noise of increasing magnitude, the model is exposed to a broader range of challenging variations, thereby improving its generalization and robustness to overfitting. Extensive experiments conducted across diverse real-world environments demonstrate that ProFi-Net significantly outperforms conventional prototype networks and other state-of-the-art few-shot learning methods in terms of classification accuracy and training efficiency.

Problem

• Overcoming limited training data in WiFi gesture recognition
• Enhancing feature discrimination via attention mechanisms
• Improving generalization with curriculum-based data augmentation

Innovation

• Prototype-based metric learning with feature attention
• Curriculum-inspired data augmentation on query set
• Dynamic Euclidean distance refinement via discriminative features