Improving Gesture Recognition with Thermal Sensors in the Presence of Body Heat Signatures

I am building a gesture recognition based on a thermal sensor.
My classification results are much worse when hand gestures are performed and the thermal sensor is “seeing” the body of the person too.
I tried to fix this by creating a class that can identify the background heat signature of someones body, but now my model only classifies the body because it covers more of the sensor than any hand gesture.
The low resolution of 8x8 pixels makes this much harder.
Is there a way to get better results?

I had a similar question before after which I implemented a filter. While that made things much better in perfect conditions, I still cannot figure out how to solve this current issue.

Project ID: 290646

Hi @rwahidi

Given the complexity of your challenge, and the nature of your project, as was suggested by Louis in the previous post it may be beneficial to consider our enterprise-tier solution features where you can access more advanced tools, resources, and ml solution expertise. An enterprise meeting with Edge Impulse could provide you with:

  • Access to specialized machine learning expertise and consultation.
  • Advanced tools and features not available in the standard platform.
  • Customized solutions tailored to the specific needs of your project.

Would you like assistance in setting up a meeting with one of our team to explore these options in more detail?



Hi @Eoin

Thank you for the offer. Since my project does not serve any practical use right now I do not need an enterprise meeting.