PPGWeaver: Diffusion-Augmented Models for Real-Time Heart Rate Estimation on Microcontrollers

Vinayak Narasimhan, Raimi Shah, Shubhankar Agarwal, Sai Shankar Narasimhan, Shailabh Kumar, Sang Kyu Kim, Sandeep Chinchali, Radwanul Hasan Siddique

Confirmation: I have read and agree with the IEEE BSN 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: Heart Rate Estimation, Photoplethysmography, PPG, Synthetic Data, Diffusion Models, Model Pruning, Edge AI, Temporal Convolutional Networks, Microcontrollers, Wearables

TL;DR: Tiny models for real-time heart rate estimation on microcontrollers trained using synthetic time-series data generated by a conditional diffusion model

Abstract: We present a compact, deployable heart rate (HR) estimation system using photoplethysmography (PPG) and inertial measurement unit (IMU) data, combining TimeWeaver, a conditional diffusion model for metadata-aware synthetic augmentation, with progressive structured pruning of Temporal Convolutional Networks (TCNs). Our smallest model, with 1.56k parameters, achieves a mean absolute error (MAE) of 4.92 BPM on the PPG-DaLiA dataset and supports real-time inference (<40 ms latency) on a 64 MHz ARM Cortex-M4F microcontroller (MCU) without requiring quantization. Synthetic data conditioned on subject metadata, HR, and activity type significantly enhances model generalization, enabling pruned models to match or exceed the accuracy of larger baselines, achieving over a 23% improvement compared to training on real data alone. Our work establishes a new Pareto frontier for real-time, on-device HR monitoring using diffusion-augmented training and sub-2k parameter models.

Track: 3. Signal processing, machine learning, deep learning, and decision-support algorithms for digital and computational health

NominateReviewer: Radwanul Hasan Siddique

Submission Number: 83