Towards Characterizing Knowledge Distillation of PPG Heart Rate Estimation Models
This work addresses the need for efficient heart rate monitoring on edge devices for health applications, but it is incremental as it builds on existing distillation methods without introducing new paradigms.
The paper tackled the problem of deploying large pre-trained PPG heart rate estimation models on wearable devices with strict memory and latency constraints by exploring knowledge distillation to create smaller models, finding that decoupled knowledge distillation (DKD) achieved the best performance with a mean absolute error of 2.1 bpm on a benchmark dataset.
Heart rate estimation from photoplethysmography (PPG) signals generated by wearable devices such as smartwatches and fitness trackers has significant implications for the health and well-being of individuals. Although prior work has demonstrated deep learning models with strong performance in the heart rate estimation task, in order to deploy these models on wearable devices, these models must also adhere to strict memory and latency constraints. In this work, we explore and characterize how large pre-trained PPG models may be distilled to smaller models appropriate for real-time inference on the edge. We evaluate four distillation strategies through comprehensive sweeps of teacher and student model capacities: (1) hard distillation, (2) soft distillation, (3) decoupled knowledge distillation (DKD), and (4) feature distillation. We present a characterization of the resulting scaling laws describing the relationship between model size and performance. This early investigation lays the groundwork for practical and predictable methods for building edge-deployable models for physiological sensing.