Addressing the issue of traditional speech collection and recognition being susceptible to environmental noiseA flexible graphene sensor is proposed to detect human vocal vibration signals. Firstly, a speech detection sensor with a cylindrical micro surface structure substrate was prepared using chemical vapor deposition (CVD) and imprinting techniques, greatly improving the conformal coating coverage and sensitivity of the sensor. In the frequency range of 200-2500 Hz, the average voltage gain of the sensor is about 48dB, which basically covers the human speech frequency.

On this basis, researchers conducted bilingual testing (in both Chinese and English). All the data obtained indicate that the graphite voice sensor has sufficient sensitivity to extract features of sound waves. Meanwhile, the proposed cylindrical micro surface structure reduces the possibility of random fracture of graphene layers. In addition, speech signals collected through microphones and flexible graphene voice detection sensors are used to train neural networks. The recognition accuracy of the dataset with mixed vocal cord speech signals is 75.9%. Compare and verify that the signals detected by the sensors have sufficient feature information to complete the speech recognition task.