Researchers use machine learning to create fabric-based touch sensors

New research from North Carolina State University combines three-dimensional embroidery technology with machine learning to create a fabric-based sensor that can control electronic devices through touch.The paper will be published in a magazine device.

As interest in the field of wearable electronics grows and new features are added to clothing, embroidery-based sensors or “buttons” that can control those features become increasingly important. This sensor is embedded in the clothing's fabric and can activate and control electronic devices, such as mobile apps, with just a touch.

This device consists of two parts. The embroidered pressure sensor itself and a microchip that processes and distributes the data collected by the sensor. The sensor is triboelectric. This means that the charge generated by the friction between multiple layers is used to power the sensor itself. It is made from a thread of two triboelectric materials, one positively charged and one negatively charged, and is incorporated into a conventional textile fabric using an embroidery machine.

Long Yin, the study's corresponding author, said the three-dimensional structure of the sensor is important for its correct use.

“Because the pressure sensor is triboelectric, we had to have two layers with a gap between them. Since we are using embroidery, which is typically two-dimensional, that gap was one of the difficult parts of the process. “Embroidery is a decorative technique. It's a fabric,” he said. “It's difficult to create three-dimensional structures like this. By using spacers, we were able to control the gap between the two layers, which allowed us to control the output of the sensor.”

Data from the pressure sensor is sent to a microchip, which converts the raw input into specific instructions for the connected device. Machine learning algorithms are key to making this happen smoothly, Yin said. The device must be able to distinguish between gestures assigned to different functions and ignore unintended inputs that may arise from the normal movement of the fabric.

“Sometimes the data that the sensors get is not very accurate, and this can happen for all kinds of reasons,” Yin said. “In some cases, the data may be affected by environmental factors such as temperature or humidity, or the sensor may accidentally touch something. Using machine learning, we can recognize such things. Machine learning also allows this very small device to perform a variety of functions because it can recognize different types of inputs. Masu. ”

The researchers demonstrated this input recognition by developing a simple music-playing mobile app that connected to the sensor via Bluetooth. They designed six functions for the app: play/pause, next song, last song, volume up, volume down, and mute, each controlled by a different gesture on the sensor. The researchers were able to use the device for several other functions, including setting and entering passwords and controlling video games.

Yin said the idea is still in its infancy because existing embroidery techniques cannot easily handle the type of materials used to create the sensors. Still, the new sensor represents another piece of the wearable electronics puzzle in development, and is sure to continue to generate interest in the near future.