Published Paper Details:

Real-time multimodal data classification has become increasingly important in applications such as healthcare monitoring, intelligent surveillance, human activity recognition, and smart environments. However, effectively capturing both spatial and temporal dependencies from heterogeneous data sources remains a significant challenge. This paper proposes a hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) framework for real-time multimodal data classification. The proposed model leverages CNN layers to extract high-level spatial features from input modalities such as images or sensor signals, while LSTM networks are employed to model temporal dependencies and sequential patterns within the data. A feature fusion mechanism is introduced to integrate multimodal representations, enabling robust and context-aware classification. The architecture is optimized for low-latency inference to support real-time deployment in edge-based systems. Extensive experiments are conducted on benchmark multimodal datasets to evaluate performance in terms of accuracy, precision, recall, F1-score, and computational efficiency. The results demonstrate that the proposed hybrid CNN-LSTM framework outperforms conventional single-model approaches, achieving improved classification accuracy while maintaining real-time processing capability. The proposed system shows strong potential for deployment in dynamic, real-world environments where multimodal data streams are continuously generated.