High‐Performance Flexible 2D Bi 2 O 2 Se Piezoelectric Energy Harvester Enabled by Structural Engineering Strategy for Silent Speech Recognition

ABSTRACT Two‐dimensional piezoelectric Bi 2 O 2 Se has attracted extensive attention in flexible self‐powered electronics. However, the application field is severely limited by insufficient output performance caused by inherent stress concentration and dissipation in conventional device architectures. Herein, we systematically characterize the intrinsic piezoelectric and ferroelectric properties of 2D Bi 2 O 2 Se nanosheets, develop a high‐performance flexible piezoelectric energy harvester (PEH) via rationally tailoring the thickness matching between the piezoelectric functional layer and flexible substrate to optimize stress distribution. This structural engineering strategy effectively alleviates stress concentration and suppresses dissipation, boosting the peak‐to‐peak open‐circuit voltage to 291 mV, 12.13 times higher than that of non‐optimized devices. The optimized PEH exhibits an outstanding sensitivity of 324.15 mV/strain(%), a fast response time of 21 ms, and a high signal‐to‐noise ratio of 40 dB, favoring conversion of laryngeal muscle movements into voltage signals. The voltage signals are accurately recognized by the microcontroller with a dual‐channel Micro‐Inception 1D classifier, enabling real‐time recognition of common speech with an accuracy of 91.03%. The recognition results are subsequently converted into intelligible speech via an audio module. This work provides a novel approach to enhance the output performance of 2D PEH and open up a brand‐new solution for the intelligent speech application.