Optimizing dictionary learning parameters for solving Audio Inpainting problem

Abstract

Abstract-Recovering missing or distorted audio signal samples has been recently improved by solving an Audio Inpainting problem. This paper aims to connect this problem with K-SVD dictionary learning to improve reconstruction error for missing signal insertion problem. Our aim is to adapt an initial dictionary to the reliable signal to be more accurate in missing samples estimation. This approach is based on sparse signals reconstruction and optimization problem. In the paper two staple algorithms, connection between them and emerging problems are described. We tried to find optimal parameters for efficient dictionary learning.

Dictionary Learning for Sparse Audio Inpainting

“…Note that we are not aware of any other dictionary learning techniques for audio inpainting for the medium gap setting. There are several related dictionary learning contributions for audio inpainting but their focus is on short gaps or on audio declipping [30], [37], [38]. We also want to emphasize that we cannot simply use out-of-the-shelf dictionary learning methods like K-SVD [39], MOD [40], or others [41], because we require our learned dictionary to satisfy specific structural properties, see Subsection V-A.…”

Late pregnancy analysis with Yunban’s remote fetal monitoring system

“…Adler et al 8 first applied the sparse representation theory in image restoration to audio repair and proposed a sparse representation of the audio repair algorithm framework. In view of the shortcomings of fixed dictionaries, Mach and Ozdobinski 9 used K-singular value decomposition (SVD) dictionary training methods to correct audio. The dictionary training method is essentially a solution to minimize formula (6)…”