This project comes from the following papers:
[1] X. Lu, S. Sun, K. Liu, J. Sun, and L. Xu, “Development of a Wearable Gesture Recognition System Based on Two-Terminal Electrical Impedance Tomography,” IEEE J. Biomed. Health Inform., vol. 26, no. 6, pp. 2515–2523, Jun. 2022, doi: 10.1109/JBHI.2021.3130374.
[2] J. Sun, X. Lu, S. Sun, R. Wang, and Y. Xie, “Application of Multi-channel Impedance Measurement Device in Teaching of ‘Signal Analysis and Processing,’” in 2021 2nd International Conference on Artificial Intelligence and Education (ICAIE), Dali, China, Jun. 2021, pp. 616–620. doi: 10.1109/ICAIE53562.2021.00136.
The device has already been verified in controlling the PowerPoint (PPT) to turn pages and jump to the outline, etc. In the future, the recognition of more gestures with more complex or subtle motion changes will be investigated, so the device can be applied in more applications mentioned in Section I [paper 1], especially in some outdoor scenes, such as sign language recognition of traffic police and construction workers in harsh environments (e.g., crowded road or rainy day, etc.).
Here are some videos recording the iterations of the project:
Video1: Prototype machine for gesture recognition - OneDrive-URL
Video2: Earlier version of software - OneDrive-URL
Video3: Real-time tomography of a goldfish swimming in a fish tank - OneDrive-URL
Impedance changes during gesture recognition include changes of contact impedance between electrode-skin and changes of internal impedance in the wrist section.
As shown in following figure, network A is the electrode-skin model describing the contact impedance. The values of R_Ele, C_Ele and R_Glu depend on the contact state (e.g., different contact areas and pressures) between electrodes and skin. Network B is the basic unit of the impedance network in the wrist section. The values of R_Omega and C_Omega depend on the type (e.g., bones, muscle, and blood vessels) and physiological state (e.g., muscle relaxation, contraction, and edema) of the biological tissue unit at that position.
The muscle and skin bulge or twist to varying degrees with different gestures, leading to changes of contact state between electrode-skin and changes of parameters in network A. Besides, gesture changes also cause the movement of biological tissues in the wrist section, resulting in changes of parameters in network B. The changes in the impedance network A&B usually cause the different conductivity distribution in the wrist section, which can be reconstructed by analyzing the impedance changes at the boundary of the wrist section. Thus, different gestures can be recognized by using our EIT system.
The following figures show the customized circuit board. The size of the circuit board is 6×4.7cm² with a cost of approximately $30. The microcontroller unit (MCU) in the system adopts a STM32F103C8T6 chip. The excitation signal is generated by a direct digital synthesizer (DDS) chip AD5933, which produces a voltage signal with a frequency of 20kHz and an amplitude range of 2V p-p. The DDS chip also integrates an ADC and a digital signal processor (DSP), which is used to collect response signals and calculate impedances. Two 16-to-1 multiplexers CH74HC4067 enable the multiplexing of the excitation and acquisition of response signal on any two electrodes. Data is transmitted to the PC side via a USB cable or wirelessly transmitted to the PC side with a Wi-Fi or Bluetooth module.
