An Expert System for Identification of Domestic Emergency based on Normal and Abnormal Sound
Swati Shilaskar, Shripad S. Bhatlawande, Aditya Vaishale, Prapti Duddalwar, Aditya Ingale · 2023
The paper describes abnormal sound identification for normal and abnormal event class. In this proposed methodology, an aggregate of regular and abnormal sounds is proposed using topological data analysis. As it classifies the different sounds using various algorithms namely Random Forest, Support Vector Machine (SVM), Decision Tree and K-Nearest Neighbors (KNN). It detects distinct events based on the dataset. This identification will help to classify which sound is safe or unusual in indoor environment. Events like fall, medical issues of elderly people, attack, hostage situation, child abuse, irregular human activities, etc. give rise to unusual sound. Proposed expert systemidentifies glass breaking, gunshot, stabbing, screaming, shouting, unexpected silence etc. as abnormal activity. The best accuracy offered by machine learning model is 98% using KNN.