Multi-mode Study of Deep Learning Applications in Acoustic Signal Processing

摘要

This proposed an intelligent acoustic signal processing system based on deep neural networks (DNNs). Employing deep learning algorithms, our system can perform acoustic scene classification, audio tagging and (rare) sound event detection, achieving a high performance in terms of accuracy, error rate and so on. An attracting feature of the system is that it integrate four working modes together, which can do parallel processing of an audio record and give results from different aspects.