2018/10/30 by Pablo A. Alvarado, Alvarado, Pablo A., Mauricio A. Álvarez +3
Computer Science · Engineering · Mathematics · Physics and Astronomy · #Gaussian Processes and Bayesian Inference #Scientific Research and Discoveries #Speech and Audio Processing #cs.LG #cs.SD #eess.AS #eess.SP #stat.ML
paper · pdf · doi:10.48550/arxiv.1810.12679
Paper submitted to the 44th International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019. To be held in Brighton, United Kingdom, between May 12 and May 17, 2019
arxiv created 2018/11/21 · arxiv updated 2018/11/22
Gaussian process (GP) audio source separation is a time-domain approach that circumvents the inherent phase approximation issue of spectrogram based methods. Furthermore, through its kernel, GPs elegantly incorporate prior knowledge about the sources into the separation model. Despite these compelling advantages, the computational complexity of GP inference scales cubically with the number of audio samples. As a result, source separation GP models have been restricted to the analysis of short audio frames. We introduce an efficient application of GPs to time-domain audio source separation, without compromising performance. For this purpose, we used GP regression, together with spectral mixture kernels, and variational sparse GPs. We compared our method with LD-PSDTF (positive semi-definite tensor factorization), KL-NMF (Kullback-Leibler non-negative matrix factorization), and IS-NMF (Itakura-Saito NMF). Results show that the proposed method outperforms these techniques.