Bayesian Restoration of Audio Degraded by Low-Frequency Pulses Modeled via Gaussian Process
Federal University of Rio de Janeiro (UFRJ)（里约热内卢联邦大学）；Rio de Janeiro State University（里约热内卢州立大学）
Hugo Tremonte de Carvalho, Flávio Rainho Ávila, Luiz Wagner Pereira Biscainho
备注：
14 pages, 7 figures, 4 tables. Submitted to IEEE Journal of Selected Topics in Signal Processing - Special Issue "Reconstruction of audio from incomplete or highly degraded observations"
A common defect found when reproducing old vinyl and gramophone recordings with mechanical devices are the long pulses with significant low-frequency content caused by the interaction of the arm-needle system with deep scratches or even breakages on the media surface. Previous approaches to their suppression on digital counterparts of the recordings depend on a prior estimation of the pulse location, usually performed via heuristic methods. This paper proposes a novel Bayesian approach capable of jointly estimating the pulse location; interpolating the almost annihilated signal underlying the strong discontinuity that initiates the pulse; and also estimating the long pulse tail by a simple Gaussian Process, allowing its suppression from the corrupted signal. The posterior distribution for the model parameters as well for the pulse is explored via Markov-Chain Monte Carlo (MCMC) algorithms. Controlled experiments indicate that the proposed method, while requiring significantly less user intervention, achieves perceptual results similar to those of previous approaches and performs well when dealing with naturally degraded signals.
Summary of The Inaugural Music Source Restoration Challenge
Independent Researcher（独立研究者）；Johns Hopkins University（约翰霍普金斯大学）；The Chinese University of Hong Kong（香港中文大学）；Sony AI（索尼人工智能）；King's College London（伦敦国王学院）
Longest Filled Common Subsequence for Song Identification from Degraded Audio via Construct--Merge--Solve--Adapt Optimization
Faculty of Natural Sciences and Mathematics, University of Banja Luka, Bosnia and Herzegovina（自然科学与数学学院，波斯尼亚和黑塞哥维那巴尼亚卢卡大学）；University of Nova Gorica, Nova Gorica, Slovenia（诺瓦戈里察大学，斯洛文尼亚诺瓦戈里察）；Artificial Intelligence Research Institute (IIIA-CSIC), UAB, Barcelona, Spain（人工智能研究 institute（IIIA-CSIC），巴塞罗那大学（UAB），西班牙巴塞罗那）；Faculty of Mathematics, University of Belgrade, Serbia（数学学院，贝尔格莱德大学，塞尔维亚）；Jožef Stefan Institute, Ljubljana, Slovenia（约泽夫·斯特凡研究所，斯洛文尼亚卢布尔雅那）；Institute of Information Sciences (IZUM), Maribor, Slovenia（信息科学研究所（IZUM），斯洛文尼亚马里博）
Token-Based Audio Inpainting via Discrete Diffusion
School of Electrical and Computer Engineering, Ben-Gurion University of the Negev（本·古里安大学电子与计算机工程学院）；Faculty of Computer and Information Science, Data Science Research Center, Ben-Gurion University of the Negev（本·古里安大学计算机与信息科学学院）；University of Haifa（海法大学）
