Classification Of Appurtenant Features To Implement Low End Ai That Classify Urban Sounds On Edge Computing Devices

Abstract

Artificial Intelligence is a way of making a machine takes its own decision. Growing demands of IoT and edge computing made deploying AI on the edge a fundamental requirement. However, edge devices cannot run all the algorithms that were built and tested on high end machines. There are various features such as MFC, MFCC’s, spectral contrast, tonnetz, CHROMA_STFT, etc. that are widely used for classification of speech and sound. Fine tuning of these features should be done depending on the use case of the application for them to run and classify on edge devices. This paper aims at classification of such appurtenant features to implement low end AI that classify urban sounds on edge computing devices.