Flexible triboelectric sensor has a broad application prospects in self-powered human-machine interface. However, due to the complexity of real environment, the output signal is easily disturbed, thus resulting in low reliability. Here, a flexible microfluidic triboelectric sensor (FMTS) was proposed by counting the number of wave peaks in output waveform. With a high transmittance of 82%, the FMTS shows bendable, twistable and conformable characteristics so that can be operated attached to skin. Based on triboelectrification and electrostatic induction between liquid stream, microchannel and interdigital electrodes, it enables to generate quantifiable voltage wave peaks. The FMTS has a maximum sensitivity of 0.418 kPa⁻¹ and a wide detection range from 2.38 to 58.12 kPa with the microchannel width of 500 µm. For demonstration, a FMTS is attached to the finger, and the number of peaks is used to determine the angle of finger bending with a resolution of 10°. The machine learning approach is employed to achieve accurate recognition of five gestures with an accuracy of 99.2%. The information can also be defined by using waveforms generated by different combinations of movements. An encoding system that recognizes eight types of information with an accuracy of up to 98.8% is finally demonstrated.

柔性摩擦电传感器在自供电人机界面方面具有广阔的应用前景。但由于真实环境的复杂性，输出信号容易受到干扰，从而导致可靠性不高。在这里，通过计算输出波形中的波峰数量，提出了一种柔性微流体摩擦电传感器 (FMTS)。FMTS 具有 82% 的高透光率，具有可弯曲、可扭曲和贴合的特性，因此可以贴在皮肤上操作。基于液流、微通道和叉指电极之间的摩擦起电和静电感应，它能够产生可量化的电压波峰。FMTS 的最大灵敏度为 0.418 kPa ⁻¹检测范围从 2.38 到 58.12 kPa，微通道宽度为 500 µm。为了演示，将 FMTS 贴在手指上，峰值的数量用于确定手指弯曲的角度，分辨率为 10°。采用机器学习的方法实现了对五种手势的准确识别，准确率达到99.2%。还可以通过使用不同运动组合生成的波形来定义信息。最终展示了一种识别八种信息的编码系统，准确率高达 98.8%。