Cross-Attention Based Multimodal Data Fusion for IoMT using Reinforcement Learning

Internet of Military Things is a domain based on IoT, which requires real-time basis data processing technique for efficient decision-making process. As military applications need to be more secure and the reliable actions need to be taken. For accurate decision making, a huge amount of information from the sensor network is required. Data processing and communication to the user end of the vast amount of data degrades the system performance. Data fusion is a technique that produces the output that represents the information from the multiple sources. This output helps in the decision-making process more efficiently. Therefore, this paper proposes the cross attention model for data fusion of multi-modal data using the Reinforcement learning algorithm in IoMT. The feature level data fusion takes place using CNN as the feature extractor. The proposed framework is simulated and obtains better AUROC and AUPRC. The proposed work also estimated the throughput which showed good performance for multimodal sensor data fusion.