21 November 2024 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Arterial blood pressure waveform (ABPW) offers comprehensive insights into cardiovascular health compared to discrete blood pressure measurements. However, accurately estimating shapes and pressure values of ABPW points in a beat-to-beat manner poses significant challenges. Current ABPW monitoring methods require invasive procedures or continuous skin contact, which are inconvenient and unsatisfactory. Thus, we propose WaveBP, the first contactless ABPW monitoring system utilizing a commercial mmWave radar, driven by the understanding that cardiac information serves as an implicit bridge between mmWave signals and ABPW based on a hemodynamics analysis model. To preserve waveform details, we design a hybrid Transformer model called mmFormer, incorporated with spatially-informed shortcuts. mmFormer enables consistent sequence-to-sequence transformations while accommodating different levels of personalization efforts. To mitigate the inherent instability of mmWave signals, we develop a beamforming-based data augmentation approach that has been empirically and theoretically proven to enhance robustness with multiple spatial observations. Additionally, we introduce a cross-modality knowledge transfer framework to fuse knowledge from cardiac modalities (ECG/PPG) with vibrations captured in mmWave reflections, improving accuracy without requiring extra deployment overhead. Extensive evaluations conducted on 43 subjects using a leave-one-subject-out setup validate that WaveBP achieves a high waveform correlation of 0.903 and exhibits a low (mean±standard deviation) error of point-level measurements at (-0.14±7.48) mmHg, which could be further reduced by subject-specific specialization. WaveBP demonstrates remarkable performance under challenging scenarios and exhibits potential for detailed cardiac estimations, as evidenced by our case studies on relative cardiac output estimation and cardiac abnormality detection.
19 December 2023 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Wearable sensor-based human activity recognition (HAR) has gained significant attention due to the widespread use of smart wearable devices. However, variations in different subjects can cause a domain shift that impedes the scaling of the recognition model. Unsupervised domain adaptation has been proposed as a solution to recognize activities in new, unlabeled target domains by training the source and target data together. However, the need for accessing source data raises privacy concerns. Source-free domain adaptation has emerged as a practical setting, where only a pre-trained source model is provided for the unlabeled target domain. This setup aligns with the need for personalized activity model adaptation on target local devices. As the edge devices are resource-constrained with limited memory, it is crucial to take the computational efficiency, i.e., memory cost into consideration. In this paper, we develop a source-free domain adaptation framework for wearable sensor-based HAR, with a focus on computational efficiency for target edge devices. Firstly, we design a lightweight add-on module called adapter to adapt the frozen pre-trained model to the unlabeled target domain. Secondly, to optimize the adapter, we adopt a simple yet effective model adaptation method that leverages local representation similarity and prediction consistency. Additionally, we design a set of sample selection optimization strategies to select samples effective for adaptation and further enhance computational efficiency while maintaining adaptation performance. Our extensive experiments on three datasets demonstrate that our method achieves comparable recognition accuracy to the state-of-the-art source free domain adaptation methods with fewer than 1% of the parameters updated and saves up to 4.99X memory cost.
21 December 2022 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Spirometry is the gold standard for evaluating lung functions. Recent research has proposed that mobile devices can measure lung function indices cost-efficiently. However, these designs fall short in two aspects. First, they cannot provide the flow-volume (F-V) curve, which is more informative than lung function indices. Secondly, these solutions lack inspiratory measurement, which is sensitive to lung diseases such as variable extrathoracic obstruction. In this paper, we present EarSpiro, an earphone-based solution that interprets the recorded airflow sound during a spirometry test into an F-V curve, including both the expiratory and inspiratory measurements. EarSpiro leverages a convolutional neural network (CNN) and a recurrent neural network (RNN) to capture the complex correlation between airflow sound and airflow speed. Meanwhile, EarSpiro adopts a clustering-based segmentation algorithm to track the weak inspiratory signals from the raw audio recording to enable inspiratory measurement. We also enable EarSpiro with daily mouthpiece-like objects such as a funnel using transfer learning and a decoder network with the help of only a few true lung function indices from the user. Extensive experiments with 60 subjects show that EarSpiro achieves mean errors of 0.20L/s and 0.42L/s for expiratory and inspiratory flow rate estimation, and 0.61L/s and 0.83L/s for expiratory and inspiratory F-V curve estimation. The mean correlation coefficient between the estimated F-V curve and the true one is 0.94. The mean estimation error for four common lung function indices is 7.3%.
14 March 2025 in IEEE Transactions on Mobile Computing
Blood pressure (BP) is a critical physiological parameter closely associated with severe diseases such as heart failure and kidney damage. Current methods either require additional or dedicated hardware, or closing touching to the devices, causing discomfort and inconvenience. Therefore, a convenient, contactless BP measurement solution is highly desired. In this work, we present Ubicon-BP, a ubiquitous, device-free, and contactless BP detection application. Ubicon-BP calculates BP based on the pulse transit time (PTT), a key feature that is medically proven correlated with BP. However, using smartphone sensors to contactless calculate PTT is non-trivial since it requires a micro-second level precision for cardiac event detection. To address this issue, we propose leveraging the acoustic sensors in smartphone to detect vibrations caused by heart valve movements, as well as camera sensors to measure finger pulses. To accurately measure heartbeat signal that are susceptible to motion, we first improve the sensing granularity of acoustic signals and then introduce the IQ-MVED model to eliminate motion interference. Furthermore, when recovering pulse signals from video signals, issues such as poor generalization performance arise. Consequently, we propose the TS-CAN and meta-learning models to obtain personalized pulse signals. Finally, we transform the extracted time-frequency features from the recovered heartbeats and pulse signals to the corresponding BP. Comprehensive testing involving 50 subjects reveal a standard deviation error of $ 4.27 \; \text{mmHg}$ for diastolic pressure and $ 6.36 \; \text{mmHg}$ for systolic pressure, respectively.