Facial Expression Recognition and Digital Images Infosecurity for Prevention Care System Application in Sudden Infant Death Syndrome Monitoring

This study presents a preventive care system for monitoring sudden infant death syndrome (SIDS), integrating digital image infosecurity and facial expression recognition (FER) technologies. For image infosecurity, a symmetric advanced encryption standard (SAES)-based scheme is introduced to ensure the secure transmission of infant images over public communication channels. In the FER, the YOLOv10 (You Only Look Once, version 10)-based classifier functions as an automatic object detection (OD) framework, enabling infant facial capture and expression recognition. Its classification mechanism categorizes expressions into four classes: normal, smile, sleep, and crying. Thus, to ensure secure SIDS monitoring, the SAES-based scheme implements robust block encryption and decryption processes to protect the privacy and security of infant images, while the YOLOv10 model enhances real-time OD, feature extraction, and pattern recognition capabilities. In experimental evaluations, a dataset of 5000 different facial expression images was self-collected and labeled with four classes of expressions. For training the YOLOv10-based classifier, the dataset was split into 60% for training (3,000 images) and 40% for testing (2000 images). The stochastic gradient descent (SGD) algorithm was employed to optimize the classifier’s model parameters for further enhancing accuracy for the intended purposes. For image infosecurity testing, the number of pixel changing rate (NPCR), unified averaged changed intensity (UACI), and structural similarity index measurement (SSIM) were employed to evaluate the performances for encryption and decryption processes, ensuring confidentiality, recoverability, and availability of transmitted images. For FER testing, the ten-fold cross-validation method was applied to evaluate the classifier’s performances, and the feasibility of the proposed method could be evaluated, achieving average Precision (%) of 91.40%, average Recall (%) of 91.15%, and average Accuracy (%) of 95.10%, respectively. The experimental results validated the feasibility of the YOLOv10-based classifier for real-time facial capture and expression recognition, demonstrating its effectiveness and reliability. Therefore, this integrated preventive care system not only safeguards sensitive message data for enhancing security levels but also facilitates accurate infant expression recognition for real-time SIDS monitoring. Hence, while assisting parents in effectively caring for their infants, it is also expected to contribute to reducing the incidence of SIDS under secure monitoring conditions.