Human-Machine Collaborative Control for Smart Homes via WhaleOptimized Iterative Learning and Multimodal Fusion

Abstract

This study proposes a multimodal collaborative control method based on an improved whale optimization algorithm and iterative learning to address the issues of insufficient multimodal fusion and poor adaptability to dynamic environments in smart home human-machine collaborative control. Firstly, by introducing a dynamic learning gain mechanism to optimize the iterative learning control algorithm, the convergence speed and tracking accuracy of the system can be improved; Secondly, a feature level and decision level fusion strategy is adopted to achieve effective fusion of speech and gesture modalities; Finally, a complete smart home human-machine collaborative control system architecture is constructed. 1) In terms of control accuracy, the research method achieves average control accuracy of 0.9212 and 0.9053 in single-device and multi-device scenarios, respectively, significantly better than particle swarm optimization genetic algorithm (0.8751) and grey wolf optimization backpropagation network (0.8346). 2) In terms of error indicators, the maximum mean absolute error (0.167) and root mean square error (0.196) are reduced by more than 50% compared to particle swarm optimization genetic algorithm (0.373/0.338) and grey wolf optimization backpropagation network (0.337/0.324). 3) In terms of system performance, the accuracy recall curve area (0.9758) is improved by 5.45%-13.68% compared to the comparison methods, the system resource utilization rate is 0.054%-0.131%, and the average response time (10.31-24.12ms) is improved by more than 30% compared to particle swarm optimization genetic algorithm (18.89ms) and grey wolf optimization backpropagation network (16.21ms). The research provides a high-precision and low latency human-machine collaborative control solution for the field of smart homes.