Single-channel blind source separation based on attentional generative adversarial network
- PDF / 5,015,462 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 43 Downloads / 218 Views
ORIGINAL RESEARCH
Single‑channel blind source separation based on attentional generative adversarial network Xiao Sun1 · Jindong Xu1 · Yongli Ma1 · Tianyu Zhao1 · Shifeng Ou2 Received: 18 June 2020 / Accepted: 3 November 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Blind single-channel source separation is a long-standing machine learning and signal processing problem. Traditional blind source separation (BSS) algorithms were proposed to solve this task utilizing multiple signal constraints. Generative adversarial network (GAN) are free from statistical constraints and samples, but the role of adversarial training in the task of BSS has not been fully demonstrated. Therefore, a new separation network model that enables to learn the known separated signal distribution from the stepwise fine estimation of the unknown mixture distribution was presented in this paper, and a self-attention mechanism was introduced to solve the problem of the blurring details of the generated image by the generator which preserves image details in the process of image separation. Compared with the existing single-channel blind source separation algorithm based on generative adversarial network-neural egg separation (NES), the detailed information of this new separation algorithm is more prominent, and the source signal in the mixed image has been separated more effectively, and has better separation performance than the classic blind source separation algorithms. Keywords Blind source separation · Single channel · Generative adversarial network · Attention mechanism
1 Introduction BSS, derived from the classic cocktail party problem (Cherry 1953), refers to the process of recovering the source signal using only the observation signal received by the one sensor when the prior information and channel transmission parameters of the source signal are unknown. "Blind" has two meanings: one is that the statistical distribution of the source is unknown, and the other is that the mixing coefficients of the channel are unknown. BSS technology has received a lot of attention in the signal processing field. Areas of application include speech separation and recognition (Seki et al. 2019), instrumental music separation (Tachioka 2019), music information retrieval (Simonetta et al. 2019), electroencephalography (EEG) signal separation (Li et al. 2015) and image processing (Dong et al. 2018). BSS is widely used in image processing, such * Jindong Xu [email protected] 1
School of Computer and Control Engineering, Yantai University, Yantai, China
School of Opto‑Electronic Information Science and Technology, Yantai University, Yantai, China
2
as image feature extraction, face recognition (Alti 2020), and moving target detection (Feng and Chun 2020). Blind image separation can restore the original appearance of the image from the contaminated image and eliminate the image quality problem caused by various factors in the acquisition process. According to the relationship between the number of source sign
Data Loading...
