Sound Source Localization and Classification for Emergency Vehicle Siren Detection Using Resource Constrained Systems

The present work introduces an intelligent system for sound source localization and classification, focusing on the critical task of detecting emergency vehicle sirens. Leveraging the STM32F411RE microcontroller and SPH0645LM4H microphones, the methodology combines generalized crosscorrelation with phase transformation for precise sound source localization, alongside Mel Frequency Cepstral Coefficients (MFCC) for robust signal processing. A neural network model, expertly trained and deployed on the microcontroller, enables real-time sound classification. The integrated printed circuit board (PCB) with eight LEDs provides visual cues, enhancing emergency vehicle detection. This work contributes significantly to sound localization, digital signal processing, and embedded systems, ultimately improving safety and responsiveness in emergency scenarios. The work is vital for enhancing road safety and emergency response efficiency by providing an intelligent system that accurately identifies and localizes the sound of emergency vehicle sirens, minimizing response times and potential hazards for all road users.