Detect activity within home using sound classification with CNN

Abstract

he technological advancement has enabled the growth and demand of Internet of Things concerning analyzing of Activities of Daily Living at Smart homes which is the key focus today. In this project, the aim is to utilize readily available Internet of Things devices and create a model that collects, analyses, and classifies sounds produced while undertaking various activities within a home. In order to achieve this, Edge Impulse platform is used since it provides the necessary tools for each stage. The platform creates an ample environment for anyone either with or without a deep understanding of Machine Learning to train and use a model. The platform enables the connection of the peripheral which in this project case was a mobile phone which is used to perform the task of collecting data and also a device to deploy the model. This project also showcases how to incorporate Python in deep Machine Learning to train a model that identifies sounds and classifies them according to the relative activity. This also shows how sound classification by use of convolutional neural network may be achieved. In the deployment stage, a Web-Assembly package is created and can be installed in the peripheral in this project it was installed in a mobile phone and accessed through the phone's browser. The creation of this package helps to understand more of how C++ and Java Script tend to work.