Research on Human-computer Interaction and Emotion Recognition based on Convolutional Neural Networks

Amidst the swift advancement of intelligent vehicles, computer vision technology has become an increasingly important core technology in the field of autonomous driving, especially in providing safety guarantees for drivers. Given the common Incidence of traffic accidents resulting from driver fatigue. The use of deep learning technology to prevent driver fatigue behavior has become an indispensable protective application in intelligent cockpit human-computer interaction. This article introduces an algorithm designed to detect drowsy driving, which relies on Convolutional neuronal networks are employed to address the challenge of fatigue during driving. The core of the research lies in constructing a model for detecting fatigue depending on YOLOv5. This model uses UTA-RLDD and YawDD datasets and extracts refined image features from video frames as available input data for the model. It can have a better detection effect on the blinking and yawning characteristics of the driver's eyelids and lips, thereby effectively determining whether they are in a fatigue state. Compared with deep learning fatigue detection algorithms, experiments have shown that its average accuracy for eye closure and yawning reaches 97.3% and 91.7%, respectively, which can effectively detect driver fatigue, prevent drivers from entering a fatigue driving state, and improve driving safety. Future research will continue to optimize and expand the model to adapt to more complex driving scenarios and make greater contributions to the development of intelligent vehicles.