Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
SOUND SOURCE SEPARATION DEVICE, SOUND SOURCE SEPARATION METHOD, AND NON-TRANSITORY COMPUTER-READABLE MEDIUM STORING PROGRAM
Document Type and Number:
WIPO Patent Application WO/2019/171457
Kind Code:
A1
Abstract:
The present invention provides a sound source separation device capable of accurately separating an individual sound source signal from a mixed signal. A sound source separation device (1) comprises: a feature extraction means (2) for extracting, using a feature extractor to which a parameter used for feature extraction is applied, a feature vector for each time-frequency bin in a spectrogram containing a mixed signal made up of a combination of a plurality of sound source signals; a clustering means (3) for classifying the extracted feature vector into a plurality of clusters; a separation means (4) for generating a sound signal for each classified cluster using a time-frequency bin included in each of the plurality of classified clusters; and a parameter update means (5) for updating the parameter of the feature extractor on the basis of a mixed signal for learning that includes an observed mixed signal.

Inventors:
KOSHINAKA TAKAFUMI (JP)
SUZUKI TAKAYUKI (JP)
KOIDA KAORU (JP)
Application Number:
PCT/JP2018/008503
Publication Date:
September 12, 2019
Filing Date:
March 06, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
NEC CORP (JP)
International Classes:
G10L21/0272
Foreign References:
JP2018502319A2018-01-25
JP2006337851A2006-12-14
JP2004126198A2004-04-22
US6430528B12002-08-06
Other References:
TAESU KIM ET AL.: "Independent Vector Analysis : An Extension of ICA to Multivariate Components", PROCEEDINGS OF THE 6TH INTERNATIONAL CONFERENCE ON INDEPENDENT COMPONENT ANALYSIS AND BLIND SOURCE SEPARATION(ICA 2006, March 2006 (2006-03-01), pages 165 - 172, XP019028810
OKUNO HIROSHI ET AL.: "Understanding two simultaneous speeches by Blind source separation", JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE SPECIAL INTEREST GROUP MATERIALS, 1ST SIG CHALLENGE, vol. 11, November 1998 (1998-11-01), pages 1 - 6, XP002951030
ANDREW SIMPSON, J.R.: "Deep Transform: Cocktail Party Source Separation via Complex Convolution in a Deep Neural Network", ARXIV, 12 April 2015 (2015-04-12), XP055315248, Retrieved from the Internet
PO-SEN HUANG ET AL.: "Deep Learning for Monaural Speech Separation", PROCEEDINGS OF THE 2004 IEEE CONFERENCE ON ACOUSTIC, SPEECH AND SIGNAL PROCESSING(ICASSP 2014, May 2014 (2014-05-01), pages 1562 - 1566, XP032617016, doi:10.1109/ICASSP.2014.6853860
Attorney, Agent or Firm:
IEIRI Takeshi (JP)
Download PDF: