Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
SOUND SIGNAL MODEL LEARNING DEVICE, SOUND SIGNAL ANALYSIS DEVICE, METHOD AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2019/163940
Kind Code:
A1
Abstract:
According to the present invention, a model, which can output an embedded vector for obtaining a set of time-frequency points at which the same sound source is dominant, can be learnt stably and in a short time. On the basis of a sound source signal spectrogram formed from a plurality of sound sources, parameters of a neural network are learnt so that the embedded vector for the time-frequency points at which the same sound source is dominant is similar to an embedded vector for each time-frequency point output from the neural network that is a CNN.

Inventors:
KAMEOKA HIROKAZU (JP)
LI LI (JP)
Application Number:
JP2019/006762
Publication Date:
August 29, 2019
Filing Date:
February 22, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
International Classes:
G10L21/0308
Foreign References:
Other References:
HERSHEY, JOHN R.: "DEEP CLUSTERING: DISCRIMINATIVE EMBEDDINGS FOR SEGMENTATION AND SEPARATION", 2016 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), March 2016 (2016-03-01), XP032900557
NAMATAME, TAKAHIRO: "Pitch recognition of multi-tone signal by the two-dimensional convolutional network with all-layer gate", IPSJ SIG TECHNICAL REPORTS (SLP), 13 February 2018 (2018-02-13)
KATO, NAOKI ET AL: "Deep Metric Learning for Video-Based Person Re-Identification by Convolutional Neural Networks", THE 23RD SYMPOSIUM ON SENSING VIA IMAGE INFORATION SSII 2017, June 2017 (2017-06-01)
Attorney, Agent or Firm:
TAIYO, NAKAJIMA & KATO (JP)
Download PDF: