FeuNet: cross-user gesture recognition and fingertip force estimation network based on feature enhancement and U-Net

Surface electromyography (sEMG)-based gesture recognition and fingertip force estimation are widely applied, but cross-user generalization remains challenging. This paper proposes FeuNet, a deep learning network integrating a feature enhancement module and a sEU-Net architecture for sEMG representation and shallow feature mapping. Furthermore, an adaptive loss function balances the optimization of both tasks. With calibration using 50% of the new user's data, FeuNet achieved 78.97% recognition accuracy, alongside a 5.51% normalized root mean square error and 67.12% coefficient of determination. Compared with existing methods, FeuNet improves both cross-user performance and calibration efficiency.