SeamPose PCB

SeamPose PCB

Research Project — UIST 2024

A printed circuit board for upper-body pose estimation through capacitive sensing.

Key Features

  • • 8 conductive seams for capacitive sensing
  • • Real-time upper-body pose estimation
  • • Integration with vision-based ground truth
  • • Machine learning model for pose prediction
View Paper (UIST 2025)

Overview

Capacitive Sensing PCB layout

PCB closeup the capacitive sensing circuit.

Capacitive Sensing

I designed a PCB for 8 conductive seams that enabled capacitive sensing for upper-body pose tracking. This PCB was made in EasyEDA and used two FDC2214 capacitive sensing chips from Texas Instruments connected to a Seeed Studio XIAO nRF52840 microcontroller.

A smaller version of this PCB was later designed in KiCAD using a ball-grid-array (BGA) nRF52840 chip as part of a modular SeamPose system, however, we did not end up moving forward with the project.

System overview of data collection

System overview of setup

Data Collection

The capacitive signals from conductive seams was combined with vision-based ground truth to create a dataset for training a DL model. Movement in the participant caused changes in capacitance that was detected by the PCB, sent to the XIAO nRF52840 microcontroller, which sent the data over bluetooth to a laptop.

Capacitive sensor signals

Signals corresponding with poses

Machine Learning

We used a CNN model to accurately predict upper-body pose from capacitive sensor data using the vision-based ground truth. I was not part of the ML part of this project and questions about that should be directed to the corresponding author, Catherine.

Demonstration Video