Monitoring minuscule mechanical signals, both in magnitude and direction, is imperative in many application scenarios, e.g., structural health monitoring and robotic sensing systems. However, the piezoelectric sensor struggles to satisfy the requirements for directional recognition due to the limited piezoelectric coefficient matrix, and achieving sensitivity for detecting micrometer-scale deformations is also challenging. Herein, we develop a vector sensor composed of lead zirconate titanate-electronic grade glass fiber composite filaments with oriented arrangement, capable of detecting minute anisotropic deformations. The as-prepared vector sensor can identify the deformation directions even when subjected to an unprecedented nominal strain of 0.06%, thereby enabling its utility in accurately discerning the 5 μm-height wrinkles in thin films and in monitoring human pulse waves. The ultra-high sensitivity is attributed to the formation of porous ferroelectret and the efficient load transfer efficiency of continuous lead zirconate titanate phase. Additionally, when integrated with machine learning techniques, the sensor’s capability to recognize multi-signals enables it to differentiate between 10 types of fine textures with 100% accuracy. The structural design in piezoelectric devices enables a more comprehensive perception of mechanical stimuli, offering a novel perspective for enhancing recognition accuracy.

在许多应用场景中，监测微小的机械信号（包括幅度和方向）是必不可少的，例如结构健康监测和机器人传感系统。然而，由于压电系数矩阵有限，压电传感器难以满足方向识别的要求，并且实现检测微米级变形的灵敏度也具有挑战性。在此，我们开发了一种由锆钛酸铅-电子级玻璃纤维复合丝组成的矢量传感器，具有定向排列，能够检测微小的各向异性变形。即使受到前所未有的 0.06% 的标称应变，所制备的矢量传感器也可以识别变形方向，从而使其能够准确识别薄膜中 5 μm 高的皱纹和监测人体脉搏波。超高灵敏度归因于多孔铁驻极体的形成和连续锆钛酸铅相的高效负载传递效率。此外，当与机器学习技术集成时，传感器识别多信号的能力使其能够以 100% 的准确率区分 10 种类型的精细纹理。压电器件的结构设计能够更全面地感知机械刺激，为提高识别精度提供了新的视角。