Ultrathin crystalline-silicon-based strain gauges with deep learning algorithms for silent speech interfaces

Abstract

A wearable silent speech interface (SSI) is a promising platform that enables verbal communication without vocalization. The most widely studied methodology for SSI focuses on surface electromyography (sEMG). However, sEMG suffers from low scalability because of signal quality-related issues, including signal-to-noise ratio and interelectrode interference. Hence, here, we present a novel SSI by utilizing crystalline-silicon-based strain sensors combined with a 3D convolutional deep learning algorithm. Two perpendicularly placed strain gauges with minimized cell dimension (<0.1 mm2) could effectively capture the biaxial strain information with high reliability. We attached four strain sensors near the subject’s mouths and collected strain data of unprecedently large wordsets (100 words), which our SSI can classify at a high accuracy rate (87.53%). Several analysis methods were demonstrated to verify the system’s reliability, as well as the performance comparison with another SSI using sEMG electrodes with the same dimension, which exhibited a relatively low accuracy rate (42.60%).

Machine Learning-Enhanced Flexible Mechanical Sensing

“…Firstly, it is often necessary to collect large amounts of diversified and rigorously vetted training data from the sensing systems to ensure a high prediction accuracy of ML model, which is a tedious and time-consuming process. For most organic material-based flexible sensors which possess intrinsic device-to-device variation and poor long-term stability, great difficulties are added in combing ML algorithms since the repeatability is directly related to model training [ 172 ]. Therefore, smarter ML algorithms need to be developed for simplified training steps and the sensor performance (especially for stability and uniformity) should be improved.…”

“…Lin et al developed a pressure sensor of resistive type to detect the throat movements of saying different instructions without the real sounds coming out, and CNN was adopted to recognize the recorded signals [ 173 ]. Recently, Yu et al developed a silent speech interface by using crystalline silicon-based strain sensors on the face, and combined a CNN algorithm to realize the recognition of 100 words at a high accuracy rate (87.53%) [ 172 ]. Lee et al proposed unique sEMG sensors on the jaw and face to collect from three muscle channels and finally realized silent speech recognition of simple instructions (Fig.…”

Fabrication of gold-doped crystalline-silicon nanomembrane-based wearable temperature sensor

“…Silicon, a widely used representative inorganic material with high reliability, has oxidation resistance property and fast response time for the active electronic material, while conventional organic materials are vulnerable to oxidation and have slow response times. 7 , 8 , 9 , 10 Meanwhile, a doping process that injects impurities can modulate the properties of silicon, such as temperature coefficient of resistance (TCR), piezo-resistance and electrical conductance. Ⅲ-Ⅴ atoms (boron, arsenic, phosphorous, etc.)…”

Lantern‐Inspired On‐Skin Helical Interconnects for Epidermal Electronic Sensors

“…[1,10,11] For the practicality of skin electronic sensors, many outstanding and imaginative epidermal electronic systems have been reported, such as facial expression recognition by surface electromyography (sEMG) or facial skin strain, hand gesture recognition by sEMG or finger skin strain, heart rate monitoring by photoplethysmography or skin strain, deep-tissue hemodynamics monitoring by ultrasonic array and so on. [4,[12][13][14][15][16][17][18] To obtain the strongest target physiological signal, the epidermal sensors will be attached to special skin parts (such as facial skin, finger skin, neck skin, etc. ), and their processing circuit boards are usually placed in the inconspicuous parts of the body (such as behind the ear, on the wrist, on the anterior midline under the clothes, etc.…”