ECHOSHIELD: 'Turning Sound into Safety' An AI-Powered Sound Detection and Safety Alert System

Criminal incidents such as assault, robbery, and other violent activities pose serious risks to individuals, particularly those who are alone in isolated areas during late hours. Many of these threats are accompanied by distinctive sounds, which can serve as crucial indicators for early detection. Existing security measures often struggle with inefficiencies, such as delays in identifying threats and inaccuracies in classification. To overcome these limitations, this research introduces a software-based system designed to analyse environmental audio and detect potential dangers. The system operates entirely through smartphones, eliminating the need for extra hardware. It continuously processes ambient sound, identifies distress-related noise patterns, and promptly alerts emergency contacts via email, SMS, or WhatsApp. Audio data sourced from an open dataset is analysed using Exploratory Data Analytics (EDA) techniques and then utilized for training advanced deep learning models, including Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN), ensuring reliable classification of different sound events.