Design and Research of a Low-Cost Gesture Recognition System Based on Electrical Impedance Tomography and Pressure-based Force Myography

<p dir="ltr">This study aims to develop a low-cost hand gesture recognition system that combines Electrical Impedance Tomography (EIT) and pressure-based Force Myography (pFMG), with machine learning methods to enhance recognition accuracy. EIT, as a non-invasive imaging technique, offers advantages such as simple structure, easy implementation, and suitability for wearable devices. However, the high cost of most accurate EIT systems limits their use in affordable applications.</p><p dir="ltr">To address this, we built a low-cost EIT system using Arduino Due, AD5934, a custom amplifier, and a multiplexer. We compared two common EIT measurement methods (2-pole and 4-pole) under this setup. In an experiment involving the recognition of five different hand gestures, results showed that the 4-pole method provided more stable signals and a higher recognition accuracy, achieving about 85% accuracy compared to 74% for the 2-pole method. To further improve performance while keeping the cost low, we integrated a pressure-based force myography (pFMG) system into the EIT system. Pressure signals from the pFMG system because of the muscle deformation were collected using a 3D-printed air chamber and pressure sensor, serving as a complementary input to the EIT data. We used Support Vector Machine (SVM), Random Forest (RF), and Linear Discriminant Analysis (LDA) to classify these multimodal signals. After the fusion of the pFMG and EIT signals, the recognition accuracy improved to about 90% with SVM and RF, along with better signal stability for the 4-pole system and the same hand gestures.</p><p dir="ltr">These results demonstrate that low-cost EIT systems are feasible for gesture recognition, with the 4-pole method showing clear advantages. Combining the EIT with the pFMG further enhances accuracy and robustness while maintaining low hardware costs, offering strong potential for applications in human-machine interface, rehabilitation, and smart prosthetics.</p>