Lightweight PPG-Based Drowsiness Detection: A Multi-Domain Feature Ablation and Computational Analysis

Real-Time Drowsiness Detection Using Dual MobileNetV2 Models on Desktop and Edge Devices

Drowsiness is a leading cause of human error in transportation and in shift-based occupational work, yet delivering reliable real-time detection on affordable, resource-constrained hardware remains difficult. This study aims to develop and evaluate a vision-based drowsiness detection system that behaves consistently across a full-power desktop and a low-cost edge device. The system couples two independent MobileNetV2 transfer-learning classifiers — one for eye state (Open/Closed) and one for mouth state (Yawn/No_yawn) — with a temporal decision engine that converts frame-level predictions into microsleep and excessive-yawning alerts. Both classifiers were trained on a merged multi-source dataset (8,548 training images) and evaluated with a class-balanced protocol (186 images/class for the eye branch and 448 images/class for the mouth branch) to remove test-set imbalance bias. The decision engine was realised as two platform-appropriate pipelines that share an offline-first, retry-capable event architecture: a duration-based, two-tier hysteresis alert on a desktop application (Haar-cascade detection, H5/float32 models) and a frame-count alert designed for a Raspberry Pi 5 edge board (MediaPipe detection, quantised TensorFlow Lite models). On the class-balanced test set the eye branch reached 95.16% (H5) / 95.97% (TFLite) accuracy and the mouth branch reached 96.65% for both formats, with above-99% cross-format prediction agreement. Converting to TFLite cut model size by 73.4% (8.99 to 2.39 MB) and single-frame model inference latency roughly thirteen-fold (about 25 to 1.9 ms, measured on the development machine). Real-time desktop sessions sustained 6.9–12.0 FPS, and the Haar detector located a face in only 15.8% of off-angle frames versus 96.9–98.6% of frontal frames. This single-session result offers a preliminary, rather than definitive, indication of the detector's pose sensitivity. A lightweight dual-MobileNetV2 design with platform-appropriate detectors shows promise for delivering consistent real-time drowsiness alerts across heterogeneous hardware tiers.

Feature Selection and Performance Analysis for Heart Rate Estimation from Smartphone Photoplethysmography

Smartphone cameras enable continuous, equipment-free heart rate monitoring, yet the photoplethysmography (PPG) signals they capture are susceptible to movement, contact instability, and environmental variation that differ sharply from controlled validation settings. This study evaluated heart rate estimation from smartphone PPG signals using a feature-based approach, five classical regression models, and Recursive Feature Elimination (RFE) on the BUT PPG v2.0.0 dataset (3,888 recordings from 50 subjects, ECG reference). A total of 21 features were extracted from time, morphology, frequency, and signal quality domains, normalized using Yeo-Johnson, and evaluated under a subject-wise 5-fold cross-validation scheme. SVR with an RBF kernel achieved the best performance (MAE 9.20 bpm, RMSE 12.05 bpm). Feature selection reduced the feature count from 21 to 12 with negligible performance loss (MAE 9.19 bpm), and a Jaccard Stability Index of 0.7513 indicated that the selected subset generalizes consistently. Condition-stratified error analysis revealed that signal quality is the primary error driver, with MAE rising from 6.49 bpm on clean signals to 9.91 bpm on noisy ones, and that dynamic activities such as walking (13.69 bpm), laughing (12.74 bpm), and coughing (11.62 bpm) produce the highest errors. These findings indicate that upstream signal quality assessment is a necessary component of a reliable smartphone PPG heart rate estimation system.

LES-YOLO: a lightweight detection method for driver drowsiness in low-light conditions

Design and Implementation of Low-Cost Non-Invasive Blood Pressure Measurement Device from Photoplethysmography with Signal Quality Assessment

Cardiovascular diseases remain a leading global health concern, necessitating accurate, non-invasive blood pressure monitoring solutions. This study presents a novel photoplethysmography (PPG)-based system that integrates three key innovations: (1) a real-time signal quality assessment (SQA) model employing a one-class SVM classifier to filter unreliable PPG segments, (2) a computationally efficient two-stage neural network (1D U-Net followed by 1D MultiResUNet) for arterial blood pressure (ABP) waveform estimation, and (3) a wearable hardware platform combining a MAX30102 optical sensor and ESP32 microcontroller for portable deployment. The SQA model uses five optimized features to achieve robust motion artifact rejection. The deep learning pipeline reconstructs ABP waveforms with mean absolute errors of 4.7 mmHg (systolic) and 4.3 mmHg (diastolic), complying with the Association for the Advancement of Medical Instrumentation (AAMI) standards. Validated on 40 subjects against reference sphygmomanometer measurements, the system demonstrates 92.5% (systolic) and 95% (diastolic) classification accuracy, with subsecond latency and low-power operation. By addressing critical challenges in motion robustness, computational efficiency, and clinical validation, this work advances the practicality of cuffless BP monitoring for telehealth and resource-limited settings.

Generalizable PPG Signal Quality Assessment Using Correlation-Derived Morphological Features

Real-Time Radar Sensor Performance Enhancement via Edge Machine Learning for Driver Drowsiness Detection

This study presents an optimized, cost-effective radar sensor system for real-time, and contactless driver drowsiness detection. The system consist on a 24 GHz frequency-modulated continuous Wave radar module integrated behind the central rearview mirror for highly sensitive analysis of driver micromovements, including head tilts, and facial dynamics. The core innovation lies in the edge-deployed sensor-specific adaptation, where a custom CNN-LSTM pipeline processes complex radar returns to estimate eye opening and mouth opening metrics without visual data. This pipeline enhances the system's ability to isolate subtle motion characteristics—such as range and velocity—improving the sensing fidelity crucial for fatigue monitoring. Deployed on a resource-constrained Raspberry Pi platform, the model demonstrates robust integration and efficiency. Experimental validation on 20 subjects shows excellent performance, achieving up to 98.1% accuracy for eye blinking, and 97.9% for spontaneous yawning. This radar-based approach advances automotive sensing by providing a privacy-preserving, high-performance solution for safety-critical applications.