You Don't Need Attention: Gated Convolutional Modeling for Watch-Based Fall Detection

2026-05-18

AI Summary

This work addresses the limitations of existing wearable fall detection methods that rely on self-attention mechanisms, which incur high computational costs and struggle to precisely capture the transient impact signatures of falls. To overcome these challenges, the authors propose Gated-CNN, a lightweight dual-stream one-dimensional convolutional architecture that processes accelerometer and gyroscope signals separately. A sigmoid-gated module selectively enhances discriminative features while suppressing noise, and fuses information from both streams for final prediction. By replacing conventional attention mechanisms with structurally aligned, computationally efficient gated convolutions, the model significantly improves temporal localization of instantaneous fall characteristics and inference speed. Evaluated on five public IMU datasets, Gated-CNN achieves average F1 scores of 90%–93%, outperforming Transformer-based baselines. When deployed on the Pixel Watch 3, it attains 97% F1 score and 98% accuracy with zero missed detections.

Abstract

Existing deep learning approaches for wearable fall detection systems rely on self-attention mechanisms that impose quadratic computational overhead, distributing weights across all time steps. This global weight distribution impairs the precise localization of the brief impact signatures that characterize falls within short, fixed-length windows. To overcome this challenge, we propose Gated-CNN, a lightweight dual-stream architecture that processes accelerometer and gyroscope streams through independent one-dimensional convolutional feature extractors, followed by (i) a sigmoid gating module that selectively suppresses uninformative background activations while amplifying fall-discriminative features, (ii) a global average pooling layer that compresses each stream into a compact fixed-length descriptor, and (iii) a shared classification head that fuses both descriptors for binary fall prediction. For offline evaluation, we evaluate the model across five wrist-mounted inertial measurement unit (IMU) datasets, achieving average F1-scores of 93%, 93%, 90%, 91%, and 90% on SmartFallMM, WEDA-Fall, FallAllD, UMAFall, and UP-Fall, outperforming Transformer baselines. For real-time evaluation, we deployed the model on a Google Pixel Watch 3 and tested across 12 participants. The model achieves an average F1-score of 97% and an accuracy of 98% with zero missed falls, showing that sigmoid gating offers a more structurally aligned and computationally efficient alternative to attention for commodity smartwatch-based fall detection.

Problem

• fall detection
• wearable sensors
• self-attention
• impact localization
• computational overhead

Innovation

• Gated-CNN
• fall detection
• wearable IMU
• sigmoid gating
• attention-free