Sleep staging helps to make an accurate diagnosis of sleep disorders. Analysis of physiological sleep data helps to identify specific sleep stages. In particular, the use of multimodal data has become the key to improving the effectiveness of sleep staging. Despite the satisfactory results, multimodal models are becoming increasingly large. Due to the complexity of the calculations and resource requirements, deploying large-scale multimodal models in clinical settings is challenging. In this paper, we propose a novel multimodal knowledge distillation technique to lightweight multimodal models. Complex multimodal teacher model knowledge is transferred to lightweight student models, ensuring that performance is maintained as much as possible while the number of parameters is reduced. We validate the effectiveness of the knowledge distillation framework on publicly available sleep datasets, providing new insights into the analysis of multimodal physiological signals.