General algorithms for estimating spectrogram and transfer functions of target signal for blind suppression of diffuse noise

Nobutaka Ito 1 Emmanuel Vincent 2, 3 Nobutaka Ono 4 Shigeki Sagayama 1
2 PAROLE - Analysis, perception and recognition of speech
Inria Nancy - Grand Est, LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
3 METISS - Speech and sound data modeling and processing
IRISA - Institut de Recherche en Informatique et Systèmes Aléatoires, Inria Rennes – Bretagne Atlantique
Abstract : We propose two algorithms for jointly estimating the power spectrogram and the room transfer functions of a target signal in diffuse noise. These estimates can be used to design a multichannel Wiener filter, and thereby separate a target signal from an unknown direction from diffuse noise. We express a diffuse noise model as a subspace of a matrix linear space, which consists of Hermitian matrices instead of Euclidean vectors. This general framework enables the design of new general algorithms applicable to all specific noise models, instead of multiple specific algorithms each applicable to a single model. The more general proposed algorithms resulted in superior noise suppression performance to our previous algorithms in terms of an output signal-to-noise ratio (SNR).
Complete list of metadatas

Cited literature [14 references]  Display  Hide  Download

https://hal.inria.fr/hal-00849791
Contributor : Emmanuel Vincent <>
Submitted on : Tuesday, August 6, 2013 - 5:17:40 PM
Last modification on : Thursday, March 21, 2019 - 2:20:12 PM
Long-term archiving on : Thursday, November 7, 2013 - 5:25:08 AM

File

ito_MLSP13.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00849791, version 2

Citation

Nobutaka Ito, Emmanuel Vincent, Nobutaka Ono, Shigeki Sagayama. General algorithms for estimating spectrogram and transfer functions of target signal for blind suppression of diffuse noise. 2013 IEEE International Workshop on Machine Learning for Signal Processing, Sep 2013, Southampton, United Kingdom. ⟨hal-00849791v2⟩

Share

Metrics

Record views

952

Files downloads

423