This academic article discusses the design and implementation of a rubble analyzer probe that uses machine learning (ML) to detect human presence in the aftermath of earthquakes. The probe is designed to listen for familiar human sounds such as ‘hello’, ‘help’, coughs, and other noises within the rubble. Additionally, it provides real-time data about environmental parameters like temperature, humidity, air quality, and pressure. The proposed probe can assess the survival prospects of people trapped within the rubble based on factors such as oxygen availability and body fluids retention. The probe uses a TinyML approach and a convolutional neural network (CNN) to classify these sounds with an accuracy of 97.45%.
Publication date: 8 Nov 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2311.02087
