Comparison of Audio Event Detection Based on CNN and LSTM Neural Network

Abstract

This paper deals with the improvement of classification accuracy by comparing and analyzing two constituent methods of sound source classification model. Recently, the utilization of artificial intelligence system is increasing in the field of society. As part of this artificial intelligence system, we propose the construction of a classification system using acoustic signals of outdoor environment. However, the signal - to - noise ratio increases in the outdoor environment compared to the indoor environment, and when the acoustic signal occurs at a distance, the intelligibility of the sound source is degraded and classification becomes difficult. Therefore the proposal of this paper is a method to extract the characteristics of the acoustic signals obtained from the microphone installed in the outdoor environment and to perform the Robust time series analysis on the environmental changes in order to solve the reduction in the classification accuracy due to the lowering of the above- Based model to improve classification accuracy. For this purpose, we collect and analyze the sound sources of the environment and construct a model based on CNN and CNN-LSTM Compare. The performance comparison of classification model is done by distance environment and noise environment, and it is learned by the same DB and compared fairly to the same test set.