Multichannel audio upmixing based on non-negative tensor factorization representation

Abstract

This paper proposes a new spatial audio coding (SAC) method that is based on parametrization of multichannel audio by sound objects using non-negative tensor factorization (NTF). The spatial parameters are estimated using perceptually motivated NTF model and are used for upmixing a downmixed and encoded mixture signal. The performance of the proposed coding is evaluated using listening tests, which prove the coding performance being on a par with conventional SAC methods. Additionally the proposed coding enables controlling the upmix content by meaningful objects.

Coding-Based Informed Source Separation: Nonnegative Tensor Factorization Approach

“…• In the NMF / NTF-based methods [27], [28] the signal short-time Fourier transform (STFT) (a redundant signal representation) amplitudes are encoded by approximating them with an NMF / NTF decomposition. In [27] the STFT phase is then entropy encoded and the rate (between phase and amplitude encoding) is allocated empirically.…”

“…Besides the fact that we consider a different coding problem, the proposed approach has the following possible advantages over [27], [28]. First, we consider a probabilistic NTF applied to the modified discrete cosine transform (MDCT) or STFT of the sources.…”

“…Thus, we opt for a logarithmic compressor, followed by a scalar quantizer and an exponential expander. It is interesting to note that Nikunen et al [27], [28] use µ-law compressor and expander to quantize NTF / NMF coefficients, and the µ-law compressor also acts as logarithmic for high values. Note finally that our NTF model has a different goal than the one presented in [27], [28].…”

“…3 We also discuss how the proposed approach relates to other relevant state of the art methods such as non-negative matrix factorization (NMF) or NTF-based coding methods [27], [28], but to the best of our knowledge this is the first attempt of using NTF models with waveform coding principles.…”

Classification of Music Genres Based on Music Separation into Harmonic and Drum Components

“…The effective and reliable separation of music sounds is the key to searching musical phrases in multipitch material (Gunawan, Sen, 2012;Nikunen et al, 2012). It also forms the basis of some automatic transcription systems.…”