OpenBliSSART: Design and evaluation of a research toolkit for Blind Source Separation in Audio Recognition Tasks

Abstract

We describe and evaluate our toolkit openBliSSART (open-source Blind Source Separation for Audio Recognition Tasks), which is the C++ framework and toolbox that we have successfully used in a multiplicity of research on blind audio source separation and feature extraction. To our knowledge, it provides the first open-source implementation of a widely applicable algorithmic framework based on non-negative matrix factorization (NMF), including several preprocessing, factorization, and signal reconstruction algorithms for monaural signals. Apart from blind source separation using supervised and unsupervised NMF, we show how the framework is useful for the increasingly popular audio feature extraction methods by NMF. Furthermore, we point out a numerical optimization for NMF, and show that NMF source separation in real-time on a desktop PC is feasible with our implementation. We conclude with an evaluation of our toolkit on supervised speaker separation, demonstrating how our algorithmic framework allows to tune the real-time factors to the desired perceptual quality.

Noise robust ASR in reverberated multisource environments applying convolutive NMF and Long Short-Term Memory

“…The minimization of d 1 (7) is performed by the multiplicative update algorithm for convolutive NMF proposed by Smaragdis (2007) and Wang et al (2009), which can be very efficiently implemented using linear algebra routines employing vectorization. Note that the asymptotic complexity of this algorithm is polynomial (O(RMNP)), and linear in each of R := R (s) + R (n) , M, N, and P. All experiments for this paper were performed with the NMF implementations found in our open-source toolkit openBliSSART (Weninger et al (2011b)) to enforce reproducibility of our results.…”

Optimization and Parallelization of Monaural Source Separation Algorithms in the openBliSSART Toolkit

“…The importance of the former two parameters on separation quality has been pointed out in [24], and various previous studies clearly suggest that different cost functions maybe optimal for different source separation problems [5,30,32]. Still, to our knowledge, the trade-off between separation quality and the RTF has been rarely investigated in the light of these parameters, although the algorithms minimizing different cost functions considerably differ in the number of required matrix operations, and their complexity (cf.…”

“…Source code and demonstrations can be found at http://openblissart.github.com/openBliSSART. We have introduced openBliSSART in [32]; since then, one remarkable development has been to parallelize computationally intensive parts of the algorithms on GPUs following the Compute Unified Device Architecture (CUDA) standard. An earlier study [1] proposed the usage of CUDA for NMF, but its evaluation was limited to a single NMF algorithm and the matrix parameterization typically encountered in musical instrument separation.…”

Automatic music genre classification based on musical instrument track separation

“…It includes various source separation algorithms, with a strong focus on variants of Non-Negative Matrix Factorization (NMF). Furthermore, supervised NMF can be performed for source separation as well as audio feature extraction (Weninger et al 2017). It should be noted that openBliSSART has built-in components to separate the HAR-MONIC and DRUM instruments.…”

“…The openBliSSART application is a C++ toolbox that provides Blind Source Separation for Audio Recognition Tasks (Weninger et al 2011). Besides the basic blind (unsupervised) source separation, classification by Support Vector Machines (SVM) using common acoustic features from speech and music processing is implemented.…”