Related Experiment Video
Updated: Aug 5, 2026
07:21
Electroencephalographic Signal Acquisition Framework for Neurodiverse: A Case Study of Dolphin-Assisted Therapy
Published on: June 27, 2025
EffSCG: An Efficient Framework for Real-Time Seismocardiogram Denoising on Resource-Constrained Edge Devices
IEEE Journal of Biomedical and Health Informatics
|July 29, 2026
Summary
We optimized deep learning models for wearable devices to run on edge units, significantly reducing latency and memory usage for seismocardiogram analysis. This enables complex physiological computing directly on devices without compromising accuracy.
Area of Science:
- Biomedical Engineering
- Computer Science
- Signal Processing
Background:
- Deep learning models for wearables often rely on centralized servers due to device limitations.
- Running models on edge devices offers reduced latency and enhanced data privacy for physiological computing.
Purpose of the Study:
- To adapt complex deep learning models for resource-constrained edge devices.
- To enable on-device processing of seismocardiogram signals for cardiovascular monitoring.
Main Methods:
- Structured pruning of neural network weights.
- Quantization of weights to integer representations.
- Integration of a faster ordinary differential equation solver.
- Application of progressive distillation.
Main Results:
- Achieved a 47.2x speedup and 79.8% memory reduction.
- Maintained 92.88% of the original accuracy.
- Demonstrated applicability to diffusion-based architectures for edge deployment.
Conclusions:
- The developed optimization framework enables efficient deep learning on edge devices.
- This facilitates real-time, private physiological computing using wearable sensors.
- The approach is suitable for various diffusion-based models in resource-scarce environments.
