Time-Series at the Edge: Tiny Separable CNNs for Wearable Gait Detection and Optimal Sensor Placement

2025-11-29

AI Summary

To address the challenge of deploying Parkinson’s disease (PD) gait detection on resource-constrained wearable devices, this work proposes two ultra-lightweight 1D convolutional neural network (CNN) architectures incorporating depthwise separable convolutions and residual connections, reducing model parameters to as few as 305—approximately one-tenth of baseline models. Leveraging short-time-window triaxial accelerometer data collected from multiple body locations, the models are rigorously validated using leave-one-subject-out (LOSO) cross-validation to ensure robust generalization. Implemented on microcontrollers, inference latency remains under 10 ms, enabling real-time, on-sensor intelligent decision-making. Experimental results demonstrate state-of-the-art performance: the best-performing model achieves a PR-AUC of 94.5% and F1-score of 91.2%, surpassing conventional threshold-based methods while maintaining minimal computational overhead. This work establishes a deployable, lightweight time-series analysis paradigm for edge-enabled early PD screening.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

We study on-device time-series analysis for gait detection in Parkinson's disease (PD) from short windows of triaxial acceleration, targeting resource-constrained wearables and edge nodes. We compare magnitude thresholding to three 1D CNNs for time-series analysis: a literature baseline (separable convolutions) and two ultra-light models - one purely separable and one with residual connections. Using the BioStampRC21 dataset, 2 s windows at 30 Hz, and subject-independent leave-one-subject-out (LOSO) validation on 16 PwPD with chest-worn IMUs, our residual separable model (Model 2, 533 params) attains PR-AUC = 94.5%, F1 = 91.2%, MCC = 89.4%, matching or surpassing the baseline (5,552 params; PR-AUC = 93.7%, F1 = 90.5%, MCC = 88.5%) with approximately 10x fewer parameters. The smallest model (Model 1, 305 params) reaches PR-AUC = 94.0%, F1 = 91.0%, MCC = 89.1%. Thresholding obtains high recall (89.0%) but low precision (76.5%), yielding many false positives and high inter-subject variance. Sensor-position analysis (train-on-all) shows chest and thighs are most reliable; forearms degrade precision/recall due to non-gait arm motion; naive fusion of all sites does not outperform the best single site. Both compact CNNs execute within tight memory/latency budgets on STM32-class MCUs (sub-10 ms on low-power boards), enabling on-sensor gating of transmission/storage. Overall, ultra-light separable CNNs provide a superior accuracy-efficiency-generalization trade-off to fixed thresholds for wearable PD gait detection and underscore the value of tailored time-series models for edge deployment.

Problem

• Develop ultra-light CNNs for on-device Parkinson's gait detection
• Compare model accuracy and efficiency for wearable sensor deployment
• Identify optimal sensor placements to minimize false positives

Innovation

• Ultra-light separable CNNs for efficient gait detection
• Residual connections enhance model performance with fewer parameters
• On-device execution enables real-time edge processing