Electromyogram-strain synergetic intelligent artificial throat

Abstract

Assisting the disabled to speak is a meaningful activity and attracts much attention in the recent years. In this work, an intelligent artificial throat has been realized based on the nanomesh, which can not only emit the sound, but also detect the voice or throat vibration generated by using the electromyogram (EMG)-strain synergetic method. The Au/PVA nanomesh with good thermoacoustic effect can be used as sound source. By theorical analyzing, the porous nanomesh structure plays an important role during the sound emitting process. The Au/PU nanomesh can also be used as the strain sensor, which shows high sensitivity, large work range, and good stability. A finite element model containing six paralleled resistance layers was proposed to explain the performance of strain sensor, which reveals the nonlinear origin of the device. The Au nanomesh can be applied as the physiological electrodes whose impendence is even lower than the commercial gel electrodes, which can be used to detect the electrocardiogram and EMG signal. In addition, the nanomesh has good water permeability, stability, and conformal property with skin. Finally, a synergetic convolution neural network (SCNN) algorithm built by ResNet18 (EMG part) and two-layers CNN (strain part) is demonstrated to distinguish the transitory voice signals detected by the nanomesh strain sensor and electrodes with the high accuracy of 98.9%. This work has great potential in the language function reconstruction.