Um sistema embarcado para detecção de sirenes utilizando aprendizado profundo

Abstract

This work presents an embedded system development for siren detection using deep learning. In urban areas, emergency vehicles sirens play a role of alerting drivers and pedestrians to an emergency situation that requires a quick response. However, the acoustic insulation of vehicles and urban noise often make it difficult for sirens to be heard. This work contributes to solving this problem by using a microcontroller, audio processing techniques, and convolutional neural networks to detect sirens and take action to assist these emergency vehicles. The system achieved good performance metrics, most above 95%. This result demonstrates the feasibility of applying deep learning in embedded systems to improve traffic safety and efficiency.