![[2602.13030] Resource-Efficient Gesture Recognition through Convexified Attention](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13030] Resource-Efficient Gesture Recognition through Convexified Attention
Summary
This paper presents a novel convexified attention mechanism for resource-efficient gesture recognition in wearable e-textile interfaces, achieving high accuracy with significantly reduced parameters.
Why It Matters
As wearable technology continues to evolve, efficient gesture recognition is crucial for enhancing user interaction. This research addresses the limitations of traditional deep learning methods in power-constrained environments, providing a solution that could enable more advanced applications in smart textiles and wearable devices.
Key Takeaways
- Introduces a convexified attention mechanism for gesture recognition.
- Achieves 100% accuracy on tap and swipe gestures with minimal parameters.
- Reduces parameter count by 97% compared to conventional methods.
- Demonstrates sub-millisecond inference times suitable for real-time applications.
- Highlights the need for further validation in real-world scenarios.
Computer Science > Machine Learning arXiv:2602.13030 (cs) [Submitted on 13 Feb 2026] Title:Resource-Efficient Gesture Recognition through Convexified Attention Authors:Daniel Schwartz, Dario Salvucci, Yusuf Osmanlioglu, Richard Vallett, Genevieve Dion, Ali Shokoufandeh View a PDF of the paper titled Resource-Efficient Gesture Recognition through Convexified Attention, by Daniel Schwartz and 5 other authors View PDF HTML (experimental) Abstract:Wearable e-textile interfaces require gesture recognition capabilities but face severe constraints in power consumption, computational capacity, and form factor that make traditional deep learning impractical. While lightweight architectures like MobileNet improve efficiency, they still demand thousands of parameters, limiting deployment on textile-integrated platforms. We introduce a convexified attention mechanism for wearable applications that dynamically weights features while preserving convexity through nonexpansive simplex projection and convex loss functions. Unlike conventional attention mechanisms using non-convex softmax operations, our approach employs Euclidean projection onto the probability simplex combined with multi-class hinge loss, ensuring global convergence guarantees. Implemented on a textile-based capacitive sensor with four connection points, our approach achieves 100.00\% accuracy on tap gestures and 100.00\% on swipe gestures -- consistent across 10-fold cross-validation and held-out test evaluation -- while...
