Direct ink writing (DIW) is a widely adopted fabrication method for flexible functional materials. However, the structure formed through filament accumulation often exhibits unavoidable wave-textured patterns. Currently, limited literature focuses on the impact of texture structures in piezoelectric composite materials on the nanogenerator output performance. This study investigated the effect of surface wave-texture’s peak-to-valley (PV) value on piezoelectric output and proposed a surface PV value control method for DIW printing based on overlap rate. For piezoelectric composite materials, representative volume element (RVE) was employed to calculate the equivalent piezoelectric parameters of the composite material. The influence mechanism of the surface PV value on the nanogenerators’ output performance was analyzed through simulation, elucidating the stress distribution and electric field intensity at different PV values. The findings reveal that as the PV value of the film surface increases from 14.29 to 80.86 μm, the sensitivity of the composite material decreases from 0.832 to 0.723 V/N. Further investigation into the optimized nanogenerator demonstrated its capability to successfully detect finger-bending signals. Additionally, a machine learning model was developed to classify various human activities—including standing, non-strenuous exercise (walking), and strenuous exercise (running/jumping)—based on signals collected by the nanogenerator, achieving an impressive recognition accuracy of 97.89%. And the sensor was successfully applied in a badminton match for analysis.