Adaptive Frontends for Biosignal Recording Systems
Biosignal recording systems typically record electrical signals from the body from many channels at a high frequency. This data is then classified using machine learning approaches to create an output of interest, such as what gesture is the user attempting to create with their muscles or whether a person is sleeping
or not. Often, the classifier does not weight the data from every channel equally, and this per channel weight or importance cannot be predetermined. However, if a channel is not as important as others, it does not have to be recorded with the same fidelity as others. This work takes advantage of this fact by building frontends that can adapt their noise and linearity to save power when is is known a classifier does not weight a given channel.
Expected Graduation Date:
May, 2025
