DWMF: A Method for Hybrid Multimodal Intent Fusion Based on Dynamic Weights

In the context of human-robot interaction, the robot's ability to accurately interpret human intentions is pivotal for the efficiency and safety of subsequent interactions. However, for the elderly population, due to the decline in physical functions, robots often fail to accurately interpret the intentions of the elderly. Moreover, traditional interaction methods exacerbate the cognitive load on the elderly. This paper proposes a hybrid multimodal intent fusion method (DWMF), based on dynamic weights, to tackle this challenge. We integrate dynamic gestures, speech-to-text conversion, and real-time environmental sensing to acquire the elderly's intentions through fusion. The method's effectiveness is validated through a block building task conducted between a robot and an elderly individual. Experimental results demonstrate that our method significantly enhances the accuracy of the robot's intention recognition, thereby reducing the cognitive load imposed on the elderly during the interaction, and exhibits substantial research and application potential in the field of assistive robotics for the elderly.