Emotion-Aware Edge Computing Framework for Next-Gen Human-Machine Collaboration

Emotion-aware computing is emerging as a critical enabler of naturalistic human-machine collaboration. With the growing demand for intelligent interfaces across edge environments such as smart homes, healthcare, and wearable devices, traditional cloud-dependent emotion recognition models suffer from high latency, privacy risks, and poor adaptability. The present research introduces an original Emotion-Aware Edge Computing Framework which combines multimodal sensing, Bi-GRU + emotion inference that uses attention and real-time adaptive interaction logic that executes directly on low-power edge computers such as Raspberry Pi 4. Lightweight sensors are used to capture facial expression, Tone and physiological (GSR and heart rate) signals and are preprocessed by normalization, landmark tracking, MFCC extraction, and statistical analysis. Mixed integration methods which involve early and late convergence are used to improve recognition performance. Optimized Bi-GRU model is implemented through TensorRT and quantization with the average accuracy of 90.1, and the inference latency has been lowered to approximately 61.5 ms, and its energy consumption has been less than$1.5 ~\mathrm{J} /$inference. Its usability and responsiveness are confirmed by the user interaction studies with a high level of satisfaction (mean 9.0/10). The suggested framework realizes strong and privacy-conscious emotion detecting at the edge, transforming the meaning of affective human-machine interaction. Future directions involve the addition of more categories of emotions, cross cultural generalization, and federated learning to constantly refine the model without affecting privacy.