Computational paradigm for women and children safety using emotion recognition using hybrid convolution neural network

Ensuring the safety of women and children remains a critical societal challenge, particularly in public and digitally mediated environments, where subtle indicators of distress often emerge prior to explicit emergency situations. Recent advancements in multimodal emotion recognition and intelligent sensing technologies present a promising opportunity to develop proactive safety monitoring systems capable of identifying potential risks before escalation occurs. This work proposes a computational paradigm for women and children safety based on multimodal emotion recognition using a hybrid deep learning framework. The proposed approach integrates convolutional neural networks (CNNs) for visual emotion analysis, contextual memory networks (CMNs) for modelling conversational and temporal dependencies, and bidirectional encoder representations from transformers (BERT) for textual sentiment understanding. Decision-level fusion is employed to combine predictions from individual modalities, thereby enhancing system robustness in the presence of noisy, incomplete, or missing data. The framework is evaluated on multiple publicly available benchmark datasets, including FER-2013, MELD, IEMOCAP, SEWA, and CMU-MOSI, following standardized training and testing protocols. A comprehensive experimental evaluation is conducted using accuracy, macro-F1 score, micro-F1 score, and per-class F1 scores, along with confusion matrices to assess class-wise performance. Experimental results demonstrate that the proposed hybrid fusion strategy consistently outperforms unimodal and bimodal baseline models, achieving improved reliability and generalization across diverse datasets and real-world conversational scenarios. Furthermore, ablation studies are conducted to analyse the individual contributions of the fusion strategy, contextual modelling components, and backbone architectures. The proposed paradigm offers a scalable and interpretable foundation for proactive safety prediction systems and can be seamlessly integrated with wearable IoT devices or mobile platforms for real-time deployment. By combining multimodal perception with contextual reasoning, this work contributes toward the development of intelligent safety systems that enable timely intervention and enhanced protection for women and children.