Deep learning techniques to derive descriptions from audio signals

• Main Author: Wu, Mengkai
• Other Authors: Jagath C Rajapakse
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/138858
• Institution: Nanyang Technological University

With the rapid growth of the Internet, the amount of video and audio data is increasing sharply. With the development of big data and artificial intelligence, audio analysis and recognition technology become more important. As the audio classification requirement increases, to classify audio and generate a description, many methods have been introduced. This project uses machine learning to achieve the classification goal through building a model with Convolutional Neural Networks or other neural networks such as Recurrent Neural Networks to categorize and generate the description for the audio. This paper includes the research I have done for generating audio descriptions using different neural network models and approaches. It starts from audio data downloading, feature extraction, image generation, and classifier training to the final audio description design and implementation. In this project, after comparison on a few types of deep neural networks, we found that deep convolutional neural networks have the overall better accuracy.