High-performance flexible piezoelectric sensor assisted by machine learning for human motion and physiological signal monitoring.

Achieving a high polar β-phase fraction in fluoropolymer piezoelectric sensors is essential for optimizing the piezoelectric coefficient (d₃₃) and sensitivity, both of which are pivotal for the monitoring of weak human physiological signals and human-machine interaction. However, traditional methods for increasing the β-phase fraction mainly rely on hydrogen bonding interactions, which necessitate complex surface modification for inorganic components. In response to this challenge, we propose chemical modulation of the interfacial ion-dipole interaction, enabling effective enhancement of the polar β-phase fraction and obtaining high sensitivity of poly(vinylidene fluoride) (PVDF)-based sensors. Herein, PVDF-based piezoelectric composites were successfully fabricated by employing metal-modified Bi4Ti3O12 (BIT) as the filler. Experimental results show that the piezoelectric coefficient (d33) of PVDF/BIT-25%MN is 47 pC·N-1. This is mainly due to the interfacial ion-dipole interactions between the variable-valence ions/defects and the polymer chains, which promote the β-phase fraction. Furthermore, this piezoelectric sensor exhibits a high power density of 1580 nW/cm2 for energy harvesting and a high sensitivity of 1.83 V·kPa-1 for sensing. Owing to its high sensitivity and excellent dynamic response, the sensor reliably monitors weak physiological signals, such as pulse and throat vibrations. Furthermore, by integrating machine learning algorithms, the system enables intelligent classification and precise recognition of complex human motion signals, facilitating the development of next-generation self-powered intelligent monitoring systems with multifunctionality.