Surface electromyography (sEMG) directly reflects muscle activity and has been widely applied in areas such as prosthetics, rehabilitation, exoskeletons, and human-computer interaction. However, traditional sEMG modeling methods are often constrained by the complexity of data and computation time, making it difficult to meet real-time control requirements. This paper proposes a rapid modeling method based on sEMG, exploring its application in hand assistive devices and medical rehabilitation. By designing sEMG datasets, extracting transient features, and using deep learning algorithms (Vision Transformer and ConvNext), the method effectively reduces modeling time and improves classification accuracy. Experimental results demonstrate that this method is both efficient and accurate in real-time control applications, making it suitable for various real-time interactive systems.