Providing home care for bedridden elderly individuals is a demanding and intricate undertaking. This study introduces a novel home-based monitoring system designed for real-time facial expression analysis to enhance adaptive care for this vulnerable demographic. The system precisely evaluates pain intensity through facial expressions, subsequently triggering appropriate responses such as voice companionship, family notifications, and emergency calls. Traditional facial expression recognition systems often exhibit high misclassification rates when applied to the elderly, primarily due to their distinct facial features, including pronounced wrinkles and altered skin texture. To address this challenge, we developed a specialized facial expression dataset specifically tailored for elderly individuals. With this dataset, we engineered a deep neural network incorporating a multi-head attention mechanism, leading to significantly more accurate facial expression recognition for the elderly. This comprehensive monitoring system is further augmented with voice recognition, various sensors, and automated control devices, offering a robust and efficient solution for supporting bedridden elderly individuals within a home care environment.