关键词: 多模态交互；智能农机；语音识别；手势控制

Abstract: In response to the low adoption rate of intelligent agricultural machinery among small and medium-sized farmers and the insufficient adaptability of traditional interaction systems, this paper develops a multimodal intelligent interaction system tailored for complex field scenarios. By integrating noise-resistant speech recognition, millimeter-wave radar gesture control, and anti-glare touch technology, a collaborative interaction system is constructed with speech as the primary mode, gesture as the auxiliary mode, and touchscreen for fine-tuning. This approach addresses the issues of complex operation, high false-touch rates, and poor environmental adaptability in traditional systems. The paper innovatively introduces an automotive controller area network (CAN) bus architecture to achieve efficient multimodal data fusion and cross-layer scheduling, thereby enhancing the reliability of command transmission. Based on a modular layered architecture that integrates multi-source sensors, adaptive switching of interaction modes is realized through a dynamic priority arbitration algorithm. The system demonstrates superior stability in scenarios such as crop lodging and high-noise environments, significantly improving human-machine collaboration efficiency and providing a technical paradigm for the intelligent transformation of small and medium-sized agricultural operations. The research results hold significant practical value for promoting the standardization of agricultural machinery interaction and the popularization of intelligent equipment.

Key words: multimodal interaction; intelligent agricultural machinery; speech recognition; gesture control