911 Call Analyzer: A Vital Tool for Detecting Critical Emergencies

Abstract

Emergency response systems must be able to promptly and accurately evaluate emergency calls. We provide a machine learning- based method in this study, called the "911 Call Analyzer," to automate the process of identifying serious crises from 911 call audio recordings. Mel- frequency cepstral coefficients (MFCCs) are used by the system to extract features, and machine learning and deep learning architectures are used for classification. To forecast the urgency and severity of each emergency call, the collected features are fed into a model that has been trained on a dataset of labelled calls. We assess the 911 Call Analyzer's performance using a test dataset, and we obtain a 91% accuracy rate with RF and XG Boost model followed by SVM with 90% accuracy, CNN with 69% accuracy and lastly LSTM with 64% accuracy. These findings show how well the suggested method works to reliably identify important crises, which helps emergency dispatchers prioritize calls and allocate resources more wisely. The 911 Call Analyzer is a tool that holds great potential for improving emergency response systems' efficacy and efficiency, which will eventually benefit those who are in need. Key Words: 911 calls, MFCCs, LSTM, CNN, SVM, RF, XG Boost.