Various sensing modalities, including external and internal sensors, have been employed in research on human activity recognition (HAR). Among these, internal sensors, particularly wearable technologies, hold significant promise due to their lightweight nature and simplicity. Recently, HAR techniques leveraging wearable biometric signals, such as electrocardiography (ECG) and photoplethysmography (PPG), have been proposed using publicly available datasets. However, to facilitate broader practical applications, a more extensive analysis based on larger databases with cross-subject validation is required. In pursuit of this objective, we initially gathered PPG signals from 40 participants engaged in five common daily activities. Subsequently, we evaluated the feasibility of classifying these activities using deep learning architecture. The model’s performance was assessed in terms of accuracy, precision, recall, and F-1 measure via cross-subject cross-validation (CV). The proposed method successfully distinguished the five activities considered, with an average test accuracy of 95.14%. Furthermore, we recommend an optimal window size based on a comprehensive evaluation of performance relative to the input signal length. These findings confirm the potential for practical HAR applications based on PPG and indicate its prospective extension to various domains, such as healthcare or fitness applications, by concurrently analyzing behavioral and health data through a single biometric signal.

Sleep staging is critical for investigating sleep quality and detecting disorders. Polysomnography (PSG) remains the gold standard, but is costly and impractical for routine monitoring. This study evaluates the feasibility of a patch-type wearable device using single-lead electrocardiography (ECG) and impedance pneumography (IPG) for multi-stage sleep classification. Data from 92 patients were collected using a wearable ECG-IPG device. Preprocessing entailed bandpass filtering, segmentation into 5-min windows with 30-s overlaps, and feature extraction in time, frequency, and nonlinear domains. Three classification methods were tested and validated using 5-fold patient-independent cross-validation across 2-class (Wake, Sleep), 3-class (Wake, rapid eye movement (REM), and Non-REM), and 4-class (Wake, REM, N1, and N2) tasks. The combined approach achieved the highest accuracy in the 2-class task (accuracy: 83.6%, area under the receiver operating characteristic (AUROC): 86.0%). For 3- and 4-class tasks, feature-based methods outperformed the others, with the RCNN achieving the best F1-score (0.618 in 3-class and 0.552 in 4-class). Modality analysis revealed that IPG + R-R interval (RRI) + motion sensors provided the highest performance, with IPG and RRI identified as the most effective in sleep staging. Feature reduction using maximum relevance and minimum redundancy (mRMR) identified the top 15 features that retained 99% of the performance of the full feature set while reducing the training time by 73%. These findings highlight the feasibility of a portable ECG-IPG system for sleep staging, balancing accuracy and computational efficiency. The proposed approach has the potential to enable continuous sleep monitoring and personalized health management in real-world applications.

Lee S, et al. Digit Health Res. 2024 Jun;2(2):e3. https://doi.org/10.61499/dhr.2024.2.e3

Lee S, et al. Digit Health Res. 2025 Mar;3(1):e1. https://doi.org/10.61499/dhr.2025.3.e1