Abstract: Aiming at the insufficient piezoelectric performance of traditional polyvinylidene fluoride (PVDF)-based flexible piezoelectric sensors, a novel coaxial nanofiber piezoelectric sensor was designed and fabricated, which uses short multi-walled carbon nanotubes (CNT) and PVDF as the core layer and PVDF as the shell layer. By analyzing the correlation between the micromorphology, structural characteristics and piezoelectric output performance of the composite fibers of the sensor, the device was further assembled into a face mask for real-time collection of human respiratory signals. Combined with a convolutional neural network for signal feature extraction, classification and recognition, an identification accuracy of 97.8% was achieved. The results show that the CNT/PVDF core-shell composite fiber membrane prepared by coaxial electrospinning technology effectively improves the electromechanical conversion efficiency of the material. This experiment not only provides a feasible scheme for intelligent face masks to realize high-precision respiratory monitoring, but also serves as a teaching case to help students understand the interdisciplinary integration of professional knowledge and artificial intelligence technology.