- Build proof-of-concept prototype
- Collected ground-truth data for Machine Learning
- Fabricated speakers using a computerized embroidery machine
- Tested numerous embroidered speakers to evaluate the optimal speaker design
- Designed a custom speaker driver to drive multiple speakers simultaneously with separated audio output
Problem: The current tracking techniques is not accurate or not convient to use for outdoor activities Approach: We embroidered a set of speakers on regular clothing, using standard thread and thin copper wire, each of which emits a particular pattern of sound that a typical smartwatch can capture. By analyzing the captured sound data, the body posture can be predicted. SoundShirt can enable precise activity tracking, which can be used, for example, in health applications.
The current body movement tracking techniques usually includes smartwatch, image processing. However, these are either not accurate or not accessible to use when walking on the street. Why we need this What problem it solves Why is this better than previous approaches
In this project, I experimented with many materials for making embroidered speakers (threads, clothing materials, and coil size). Anandghan and I also designed several circuits for speakers and microphones. Since we are still working on the project, I cannot show much detail on the website. Chat with me about the project!
|First prototype - Machine learning predicting result|