Research and Application of Intelligent Driving Navigation Voice Based on Deep Learning Technology

With advancements in artificial intelligence, the Internet of Things, and sensor technology, intelligent driving-related technologies have emerged as a central focus for the global automotive industry's upgrade. Traditional in-vehicle voice systems primarily rely on rule-based engines or simple text-to-speech (TTS), which exhibit several limitations, including mechanical and monotonous prompt voices, inadequate multimodal coordination, and a lack of context awareness. These shortcomings hinder the ability of car navigation prompts to alleviate drivers' emotional anxiety during traffic jams. This paper proposes a design methodology for in-vehicle navigation voice systems based on deep learning technology, emphasizing the perception of pleasure. By integrating traffic environment data with driver emotion data and employing multimodal technology, we construct a CBGRU + graph neural network model. This model outputs navigation prompt voices that evoke pleasure and quantify the correlation between voice features (such as timbre, speech rate, emotional expression, and semantic content) and emotional perception (including pleasure and stress relief). Experimental results demonstrate that navigation prompt voices designed to evoke pleasure during traffic jams can effectively mitigate drivers' negative emotions and enhance the overall driving experience. This research on the design of car navigation prompt sounds, oriented towards the perception of pleasure, provides both theoretical and empirical support for voice interaction design in intelligent transportation systems.