Wearable ppg representations improve health monitoring with multi-scope alignment
HiPPT: Physiology-Grounded Wearable PPG Representation Learning via Multi-Scope Text Alignment
Artificial Intelligence
Summary
Wearable devices can track your heart signals using PPG sensors, but existing methods do not fully capture all the details at different time scales. The authors present HiPPT, a new way to train these devices to recognize both local pulse shapes and broader context about the wearer’s physiology. They pair over 4 million recordings with text descriptions about the pulse and user information, teaching the model to connect the waveforms with their meanings. This leads to better predictions of blood pressure, stiffness, and heart rate across many tests, and even allows the model to guess age, sex, and body metrics without extra training.
What this means in practice
- •For wearable device engineers: Design improved wearable health monitors that use multi-scope trained PPG representations for more accurate cardiovascular and physiological measurements.$Commercial implications: Enables development of wearable devices with pretrained models that improve measurement accuracy and health insights beyond current standards.
- •For health data platform developers: Build systems that aggregate PPG data with metadata to enable context-aware health analytics and personalized physiological profiling.
Authors
Jiaqi Wang, Wei Chen, Xiaoyu Li, Lexian Xia, ShuaiLong Tang, Chenran Zhang, Donglin Xie, Guangpu Zhu, Yelei Li, Yi Zhou
Abstract
Reusable wearable PPG representations should capture local pulse morphology and broader physiological regularities. Existing pretraining objectives, however, usually apply supervision at the recording or pair level, overlooking that different concepts become reliable over different scopes of waveform evidence. We propose HiPPT, a physiology-grounded framework that matches each supervision source to its appropriate evidence scope. We construct a PPG-Text corpus of over 4 million wrist-worn recordings from 202K users, pairing recordings with waveform-derived morphology and metadata-derived context descriptions. HiPPT combines record-level morphology alignment, which grounds individual recordings in pulse characteristics, with cross-recording context alignment, which aggregates recordings sharing the same metadata tuple to capture context-consistent regularities. Waveform-guided text modeling further recovers masked contextual concepts from PPG patch representations, providing fine-grained waveform-text supervision. Across six downstream tasks spanning blood pressure, vascular stiffness, physiological attributes, and cross-dataset heart-rate estimation, HiPPT achieves the best result on the primary linear-probing metrics. The aligned space also supports zero-shot profiling of age group, sex, and BMI category, individually and jointly, with performance improving as more recordings are aggregated. These results show that evidence-scope matching yields transferable and semantically accessible wearable PPG representations. Code and pretrained weights will be released.