Audio Scene Classification Under Convolutional Neural Network

Audio Scene Classification Under Convolutional Neural Network

연구 분야: Infrastructure

학회: ASENS '24: Proceedings of the International Conference on Algorithms, Software Engineering, and Network Security

Audio scene classification is a way of supporting security monitoring applications such as audio surveillance, anomaly detection, and risk management by recognizing and categorizing environmental labels in audio data. With the significant increase in the volume of audio data generated by audio-video surveillance systems, the limitations of traditional classification methods are becoming increasingly apparent. In contrast, deep learning techniques, leveraging their advantages in data feature processing and pattern recognition, have become key technologies for solving such problems. Building upon this, this paper focuses on optimizing the audio scene classification system using a Convolutional Neural Network model and delving deeper into existing dataset information without increasing additional data volume. Additionally, the network structure is adjusted without increasing the computational burden. This model approach effectively improves the recognition accuracy in specific scenarios, as evidenced by comparison analysis with a human baseline system.