forum | IKS - Details

Sie verwenden einen Browser, in dem JavaScript deaktiviert ist. Dadurch wird verhindert, dass Sie die volle Funktionalität dieser Webseite nutzen können. Zur Navigation müssen Sie daher die Sitemap nutzen.

You are currently using a browser with deactivated JavaScript. There you can't use all the features of this website. In order to navigate the site, please use the Sitemap .

Vortrag im "forum | IKS": Probabilistic Nonnegative Matrix / Tensor Factorization for Multichannel Audio Source Separation and Other Audio Inverse Problems

Dr. Alexey Ozerov, Technicolor Research & Innovation, Rennes
Mittwoch, 10. Januar 2018
17:00 Uhr
Hörsaal 4G | IKS

Dr. Alexey Ozerov, Senior Scientist in Technicolor Research & Innovation, Rennes, contributes to our IKS event series with alecture entitled:

"Probabilistic Nonnegative Matrix / Tensor Factorizationfor Multichannel Audio Source Separation and Other Audio Inverse Problems"

Nonnegative matrix factorization (NMF) and nonnegative tensor factorization (NTF) are powerful models for audio signals,and they have found many applications in audio analysis (e.g., music transcription or audio events detection) and in solvingaudio inverse problems (e.g., source separation or denoising). While originally those models were applied to approximatenonnegative spectrograms of audio signals, the introduction of NMF/NTF models with Itakura‐Saito (IS) divergence hasextended the range of their application. Indeed, this modeling has been shown equivalent to a probabilistic Gaussian modelwith structured variances of the complex‐valued short‐time Fourier transform (STFT) coefficients of the signals. This meansthat these models are Gaussian models of the signals itself, and, thanks to this Gaussianity, they allow approximating latentpower spectrograms of audio signals under various linear transforms such as filtering, summation and subsampling.

In his talk we will develop on the NMF/NTF modeling with IS divergence. We will show how, thanks to its Gaussianformulation, it can be effectively applied to various audio inverse problems including blind and supervised multichannelaudio source separation, informed audio source separation, audio declipping and compressive sampling recovery.

After the lecture, the guests have the opportunity to join an open discussion during a snack.