High‐Performance Voice Recognition Based on Piezoelectric Polyacrylonitrile Nanofibers

Abstract

Acoustic sensors show a wide application in diverse fields. However, it is still a challenge to develop highly sensitive transducing materials for the detection of low‐to‐middle decibel sounds. Herein, for the first time the remarkable property of an acoustic sensor based on electrospun polyacrylonitrile (PAN) nanofibers in the recognition of human voices is reported. The nanofiber device shows a sensitivity as high as 23 401 mV P−1 to 70 dB sound, which is much higher than most of the acoustic sensors reported in the literature. First the sounds from different musical instruments are used as a standard sound source to show the high accuracy of the nanofiber device to distinguish different sounds, e.g., from both the same and different instruments, and then people's voices are tested. Experimental results show that the device can distinguish people's voices with high resolution and the influence of background noise on speech recognition is very small. The acoustic sensor is stable and can be used in sound detection for a long time. Electrospun PAN membranes may be useful for the development of voice recognition systems for security, environmental protection, scientific research, and other high‐tech applications.

Surface Potential Tuned Single Active Material Comprised Triboelectric Nanogenerator for a High Performance Voice Recognition Sensor

“…, thus the desired acoustic output signals. The mechanoacoustic sensitivity ( S s ) of S‐TENG is calculated by the following equation [ 55–57 ] : $$\[ \begin{array}{*{20}{c}}{{S_s} = \frac{V}{P} = \frac{V}{{{P_0}{{.10}^{Lp/20}}}} = \frac{{C\Delta \varphi X/eA{\varepsilon _0}}}{{{P_0}{{.10}^{Lp/20}}}}}\end{array} \]$$where P is the sound pressure, V is the voltage output of the S‐TENG, P 0 is the reference sound pressure of 0.00002 Pa, and L p is the sound pressure level (SPL) in decibel. Figure 4b illustrating mechanoacoustic sensitivity S s of S‐TENG and a commercial electret microphone values at three different SPL, i.e., under 60, 65, and 70 dB.…”

“…By observing the ultrahigh mechanoacoustic sensitivity, wide bandwidth response, and low‐to‐middle decibel response (10–1100 Hz), S‐TENG is examined for the voice recognition. [ 57 ] It is tested with the sound of different musical instruments (such as guitar and drum) to check whether it can able to distinguish the sound of different musical instruments or not. The time domain (Figure S6a,b, Supporting Information) and frequency domain responses (Figure S6c,d, Supporting Information) indicate that different musical instruments involves different frequency, tone, amplitude, etc.…”

“…, thus the desired acoustic output signals. The mechanoacoustic sensitivity (S s ) of S-TENG is calculated by the following equation [55][56][57] :…”

Machine-Learning-Aided Self-Powered Assistive Physical Therapy Devices

“…(f) Pearson correlation coefficients calculated among the “db5” decomposed signals recorded from various musical instruments (I1, I2, I3, I4, I5, I6, I7, I8, representing piano, cello, basslap, organ, flute, guitar, trumpet, and violin, respectively). Adapted with permission from ref . Copyright 2021 Wiley-VCH GmbH.…”

“…Another investigation of voice recognition with ML assistance is based on a PENG fabricated from electrospun polyacryonitrile (PAN) nanofibers. The structure of the nanofiber sensor consists of a piece of electrospun PAN nanofiber membrane sandwiched between two Au-coated polyethylene terephthalate (PET) films as piezoelectric electrodes (Figure d) . By processing the voltage output from the sound played by the piano and trumpet, the Daubechies wavelet of order 5 (“db5”) decomposed signals from the sounds generated by the piano and the trumpet (Figure e).…”