Fatigue driving is a crucial factor in causing traffic accidents. To enhance the accuracy and reliability of determining the driver's fatigue state, automatically adjust the seat based on the judgment result, and further stimulate the driver's fatigue state to achieve the goal of safe driving, this paper focuses on the extraction of drivers’ facial and physiological characteristic data and the construction of a multimodal fusion model. Firstly, it deeply analyzes the basic theories related to the face, heart rate, and electroencephalogram (EEG), elaborates on the extraction methods of various features and their associations with the fatigue state, and introduces the applicable recognition methods. A driving simulation platform is utilized to conduct fatigue driving experiments, collect facial video, heart rate signal, and EEG signal data, and construct a fatigue driving dataset. Subsequently, a multimodal fatigue state recognition model based on BCL-SVM is proposed. The facial, heart rate, and EEG features are respectively input into Back Propagation Neural Network (BP), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) networks for preliminary prediction, then the decision fusion is carried out through the Support Vector Machine (SVM) to determine the driver's fatigue state. Finally, based on the determined result, a seat adaptive adjustment method model is proposed, providing ideas for alleviating driver fatigue and improving driving safety.