Flexible wearable vibration optical sensor for voice recognition enhanced property by a microfiber interferometer

A flexible wearable voice monitoring vibration optical sensor based on polydimethylsiloxane (PDMS) package is proposed and demonstrated, which is mainly composed of a microfiber ring as the core sensing unit, and is combined with an artificial neural network to accurately identify human voices. The sensor exhibits a vibration sensitivity of -0.0316 nm/Hz under the condition of 100 Hz to 900 Hz. In addition, taking into account the accuracy of monitoring, the matrix decoupling of axial strain (sensitivity of -0.0018 με) and temperature (sensitivity of -1.7286 nm/℃) relative to vibration is constructed by cascading an FBG to eliminate the interference between axial strain and temperature on vibration. In addition, this sensor has different voltage amplitude and response time to different speech signals, and the speech signals are collected several times in order to verify the accuracy of the sensor used for voice monitoring. The neural network model is trained and tested by using the speech waveform data obtained by the sensor, and the vocal cord vibration of different words is successfully recognized with a total recognition accuracy of 98.33%. This proposed sensor demonstrates structural simplicity, high sensitivity, and excellent flexibility, showing significant potential for applications in the field of biomedical monitoring, such as voice control, voice healthcare monitoring, and voice authentication.