Emergency Vehicle Siren Direction Detection Using Deep Learning

The siren identification of emergency rescue vehicles is important, but if it can also identify the direction, its application will be even wider. This paper proposes that three types of siren positions can be determined for emergency rescue vehicles: cars coming from the front, cars coming from behind, and ambient sounds. The proposed method is compared with the other four methods. The emergency vehicle siren in the paper is presented with the sound of an ambulance. In the test, experiments were performed with different sound intensities and different signal-to-noise ratios. The results presented in the experiments show that the proposed method CNN+BiLSTM has the highest accuracy.