No abstract

No abstract

Aiming at the problem that traditional speech acquisition and recognition are susceptible to environmental noise, this paper proposes a flexible graphene sensor to detect vocal vibration signals. First, the speech detection sensor with a cylindrical microsurface structure substrate is prepared by chemical vapor deposition (CVD) and imprint technology, which greatly improves the conformal coating cover ability and sensitivity of the sensor. In the range of 200–2500 Hz, the average voltage gain of the sensor is ∼48 dB, and this frequency range basically covers the human speech frequency. On this basis, we conducted a bilingual detection (Chinese and English). All data obtained shows that the graphite speech sensor has sufficient sensitivity to extract the characteristics of acoustic waves. At the same time, the proposed cylindrical microsurface structure reduces the probability of random fracture of the graphene layer. In addition, the speech signals collected by a microphone and the flexible graphene speech detection sensor are used to train a neural network. The recognition accuracy of the data set mixed with vocal cord speech signals is 75.9%. The comparison verifies that the signals detected by the sensor have sufficient characteristic information to complete speech recognition tasks.

No abstract