Application of a CNN-based Acoustic Classification Model to Detecting a Target Signal with Interference

In recent years, there have been several collapse accidents in Korea that have increased anxiety about disaster accidents. Several preliminary predictive studies have been conducted to prevent collapse accidents. Still, there are few studies on detecting survivors who need rescue after an accident. In this paper, we propose a convolutional neural networks-based abnormal sound classification model for detecting collapse survivors. The proposed model classifies nine types of sound, and its accuracy is 96.34 %. As an additional experiment, accuracy is measured for various SIR values to investigate the performance of this model when input signals are mixed with interfering signals.