Real Time Baby Facial Expression Recognition Using Deep Learning and IoT Edge Computing

Babies are one of the biggest groups that can't communicate verbally. Therefore, facial expressions are a good way to analyze the mood and health conditions of babies. Sleep and excessive crying disorder are common in babies that have a severe negative impact in later part of life such as behavioural problems, learning difficulties, irritability, and sometimes autism spectrum disorder (ASD). Early detection may prevent negative consequences, but that requires continuous monitoring of the facial expressions. Continuous monitoring through the classical IoT system is very expensive because the cloud infrastructure is required to run complex algorithms such as machine/deep learning. Also, an uninterrupted internet facility is required to make this system work. In this paper, a low-cost IoT edge computing and multi-headed 1-dimensional convolutional neural network (1D-CNN) model is proposed for continuous monitoring and classification of the facial expressions into happy, crying, and sleeping categories. IoT edge device locally powers the cloud computing capability, which significantly improves the security, minimizes latency and saves bandwidth costs. The performance of the proposed approach in terms of precision, recall, and f1-score is calculated and compared with the machine learning models. The proposed approach outperforms all machine learning models in all categories (happy, crying, and sleeping).