ML at the edge and in the Cloud with Azure Sphere

The Discussion

• 

  Hi,
  
  Do you have any hybrid ML examples using Azure Sphere and Cognitive Services for recognizing/verifying voices (instead of faces)?

• 

  Hi Marty - we have some explorations in this area but nothing that is posted publicly. Can I ask for more context on your goals and I can try to help?

• 

  @jamesscott,
  
  Thank you for your reply. The context of my question is to perform research and a proof of concept (using the MT3620 Azure Sphere Starter Kit from AvNet) for the following use cases in healthcare:
  
  1) An end user uses their voice to authenticate to a remote patient monitoring device in their home. Today's commercial solutions simply use a wake word where basically anyone can impersonate the end user.
  
  2) Voice based authentication in a clinical scenario where healthcare workers can have a low friction user experience for biometric based authentication into Azure Sphere (and by extension local IoT capabilities and cloud capabilities). The usage of PPE make conventional biometric authentication (face, finger) difficult to use.
  
  I am wondering if there can be a good mix of TinyML on the Azure Sphere device with more robust AI/ML capabilities in the Azure cloud when needed.
  
  Does that help provide further context?

• 

  Thanks Marty, really interesting use case!
  Thierry Moreau at OctoML has published a blog around doing wake word detection on Azure Sphere, check out: https://medium.com/octoml/bringing-tiny-ml-to-the-cloud-connected-iot-with-tvm-2ab800b065b7