基本信息
- 标题: "SampleRNN: An Unconditional End-to-End Neural Audio Generation Model"
- 作者:
- 01 Soroush Mehri (University of Montreal)
- 02 Kundan Kumar (IIT Kanpur)
- 03 Ishaan Gulrajani (University of Montreal)
- 04 Rithesh Kumar (SSNCE)
- 05 Shubham Jain (IIT Kanpur)
- 06 Jose Sotelo (University of Montreal)
- 07 Aaron Courville (University of Montreal)
- 08 Yoshua Bengio (University of Montreal)
- 链接:
- ArXiv
- Publication ICLR2017Poster
- Github
- Demo
- 文件:
In this paper we propose a novel model for unconditional audio generation based on generating one audio sample at a time. We show that our model, which profits from combining memory-less modules, namely autoregressive multilayer perceptrons, and stateful recurrent neural networks in a hierarchical structure is able to capture underlying sources of variations in the temporal sequences over very long time spans, on three datasets of different nature. Human evaluation on the generated samples indicate that our model is preferred over competing models. We also show how each component of the model contributes to the exhibited performance.