# Modular Soft Sensor Made of Eutectogel and Its Application in Gesture Recognition

## TL;DR

A new modular soft sensor made of eutectogel is developed for wearable devices, enabling easy assembly and accurate gesture recognition.

## Contribution

The introduction of a modular soft sensor unit (M2SU) with eutectogel and magnetic blocks for in situ assembly and dual functionality.

## Key findings

- The M2SU design allows for easy and reversible integration onto wearable devices.
- Gesture recognition tasks achieved high Top 3 accuracies using a 1D convolutional neural network.
- The modular design supports high-accuracy gesture recognition without re-prototyping.

## Abstract

Soft sensors are designed to be flexible, making them ideal for wearable devices as they can conform to the human body during motion, capturing pertinent information effectively. However, once these wearable sensors are constructed, modifying them is not straightforward without undergoing a re-prototyping process. In this study, we introduced a novel design for a modular soft sensor unit (M2SU) that incorporates a short, wire-shaped sensory structure made of eutectogel, with magnetic blocks at both ends. This design facilitates the easy assembly and reversible integration of the sensor directly onto a wearable device in situ. Leveraging the piezoresistive properties of eutectogel and the dual conductive and magnetic characteristics of neodymium magnets, our sensor unit acts as both a sensing element and a modular component. To explore the practical application of M2SUs in wearable sensing, we equipped a glove with 8 M2SUs. We evaluated its performance across three common gesture recognition tasks: numeric keypad typing (Task 1), symbol drawing (Task 2), and uppercase letter writing (Task 3). Employing a 1D convolutional neural network to analyze the collected data, we achieved task-specific accuracies of 80.43% (Top 3: 97.68%) for Task 1, 88.58% (Top 3: 96.13%) for Task 2, and 79.87% (Top 3: 91.59%) for Task 3. These results confirm that our modular soft sensor design can facilitate high-accuracy gesture recognition on wearable devices through straightforward, in situ assembly.