IoT-Enabled Deep Learning Framework for Smart Infant Health Monitoring

The Intelligent Baby Monitoring System integrates AI-driven cry detection, sleep tracking, and environmental safety analysis to enhance infant care. The system employs CNN-BiLSTM and Wav2Vec2.0 for cry classification, distinguishing between hunger, pain, and sleepy cries based on pitch, energy, and duration. Sleep monitoring utilizes YOLOv8, OpenPose, and LSTM to analyze sleep stages, detecting awake, asleep, restless, and distress states through body movement intensity. Anomaly detection is performed using an LSTM Autoencoder, identifying abnormal gas levels (MQ-135), temperature and humidity fluctuations (DHT22), and deviations in heart rate and SpO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> levels (MAX30102). Heart rate monitoring leverages 1D-CNN and LSTM, classifying conditions such as bradycardia (<100 bpm) and tachycardia (>180 bpm), while oxygen saturation levels below 90% indicate hypoxia. Cradle safety monitoring employs YOLOv8 and CNN-LSTM, detecting movement patterns such as excessive shaking and escape attempts, with acceleration thresholds exceeding 0.7g triggering alerts. The system is implemented on an NVIDIA Jetson Nano, ensuring real-time inference with optimized TensorRT models. Data transmission to AWS IoT Core enables cloud-based storage and remote monitoring through a mobile application. By integrating deep learning, edge AI, and cloud computing, this system offers an efficient, intelligent, and real-time solution for infant monitoring, ensuring timely alerts and enhanced caregiver response. The proposed system is evaluated across multiple performance metrics, demonstrating high accuracy in cry classification, sleep tracking, and anomaly detection, making it a robust tool for infant safety and well-being.