Single Channel Audio Source Separation using Convolutional Denoising\n Autoencoders

2017/03/23 by Emad M. Grais, Mark D. Plumbley, Grais, Emad M. +1
Computer Science · #68T01 #FOS: Computer and information sciences #H.5.5 #I.2.6 #I.4.3 #I.5 #Music and Audio Processing #Sound (cs.SD) #Speech Recognition and Synthesis #Speech and Audio Processing

paper · pdf · doi:10.48550/arxiv.1703.08019

openalex publication_date 2017/03/23 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Deep learning techniques have been used recently to tackle the audio source\nseparation problem. In this work, we propose to use deep fully convolutional\ndenoising autoencoders (CDAEs) for monaural audio source separation. We use as\nmany CDAEs as the number of sources to be separated from the mixed signal. Each\nCDAE is trained to separate one source and treats the other sources as\nbackground noise. The main idea is to allow each CDAE to learn suitable\nspectral-temporal filters and features to its corresponding source. Our\nexperimental results show that CDAEs perform source separation slightly better\nthan the deep feedforward neural networks (FNNs) even with fewer parameters\nthan FNNs.\n

Citations

• Theano: new features and speed improvements
• Convolutional Recurrent Neural Networks for Music Classification

Related

• Audio Source Separation Using a Deep Autoencoder
• Multi-Resolution Fully Convolutional Neural Networks for Monaural Audio Source Separation
• Discriminative Enhancement for Single Channel Audio Source Separation using Deep Neural Networks
• Raw Multi-Channel Audio Source Separation using Multi-Resolution\n Convolutional Auto-Encoders
• Multi-Task Audio Source Separation