A new sound forensics approximation: An automated location detection method i̇n multistorey buildings using environmental sound classification

Environmental sound classification has continued to attract research and applications designed specifically for digital forensic studies and cybercrime analysis because of its vast range of applications. [1]. Conventional ambient noise classification systems, for example, depend on hand-crafted extracted features such as MFCC, zero-crossing rate, and short-term energy. However, this might not help for position-based sound classification such as those in multi-storey buildings. This thesis is a research project aimed at resolving this issue using a centre-symmetric nonlinear pattern. We gathered ambient sounds from each floor of a multi-story structure, which we turned into a dataset and made publicly available. A substitution box (S-Box) is used to generate features in a new centre symmetric nonlinear pattern. After the feature generation technique we used, we dubbed the model CS-Lblock-Pattern. The dataset contains ten classes, one for each of the building's ten floors. Our investigations show that using SVM, we can get good classification accuracy for multi-storey hospital datasets, with an accuracy rate of 95.38%. The classification phase results clearly shows that ESC may produce effective results by applying a center-symmetric nonlinear pattern