Dual-Sensor Fusion for Real-Time Monitoring of Crystal Methamphetamine Withdrawal Symptoms Using Machine Learning and Gramian Angular Fields

Managing withdrawal from crystal metham-phetamine is challenging due to limitations in existing wearable technology, such as single sensor setups and accuracy issues. This research proposes a dual-sensor system that combines ankle and wrist sensors, enhanced by a Kalman filter and sensor fusion techniques to improve data precision and responsiveness. Data quality was enhanced using techniques like Synthetic Minority Over-sampling Technique (SMOTE) and adding Gaussian noise, and developed advanced pattern recognition abilities using Gramian Angular Field (GAF) transformations. The use of these transformations through Convolutional Neural Networks (CNN) and ConvLSTM significantly enhances the ability to detect withdrawal symptoms. The sophisticated identification methods integrated into a smartphone app built with Flutter and hosted on a FastAPI backend enable greater support for individuals recovering from drug use disorders.