With the growth of the Internet of Things and the rise of Big Data, dataprocessing and machine learning applications are being moved to cheap and lowsize, weight, and power (SWaP) devices at the edge . The field of Cyber-PhysicalMeasurements and Signature Intelligence (MASINT) makes use of these devices to exploit data in ways not otherwise possible . However, methods to train and deploy models at the . edge are limited, and models withsufficient accuracy are often too large for the edge device . Therefore, thereis a clear need for techniques to create efficient AI/ML at . the edge. This workpresents training techniques for audio models in the field of environmentalsound classification at the the edge, including training and training Transformers to classify office sounds in audio clips . Our final model outperforms the . office sounds dataset, using just over 6,000 parameters — small enough to run on a microcontroller, using over6,000 . It can outperform the .

Author(s) : David Elliott, Carlos E. Otero, Steven Wyatt, Evan Martino

Links : PDF - Abstract

Code :

Keywords : edge - models - training - devices - audio -

Leave a Reply

Your email address will not be published. Required fields are marked *