Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
SOUND DATA LEARNING SYSTEM, SOUND DATA LEARNING METHOD, AND SOUND DATA LEARNING DEVICE
Document Type and Number:
WIPO Patent Application WO/2020/071015
Kind Code:
A1
Abstract:
This sound data learning system extracts J feature quantities of a sound signal for learning, and generates N feature quantity reduction blocks that have parameters for defining whether each of the J feature quantities is in use. The sound data learning system executes a feature quantity reduction process for repeating: a preparation process for preparing a data set of k types of different patterns from the sound signal for learning; a learning process in which, for each of the N feature quantity reduction blocks, the applicable feature quantity reduction block and a portion of the data set are used for each different data structure; and an evaluation process pertaining to the use of each of the J feature quantities for the remainder of the data set. The sound data learning system selects one of the feature quantity reduction blocks as reduction information on the basis of the result of the feature quantity reduction process for each of the N feature quantity reduction blocks having k types of evaluation results.

Inventors:
IZUMI AKITOSHI
Application Number:
PCT/JP2019/033919
Publication Date:
April 09, 2020
Filing Date:
August 29, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
PANASONIC IP MAN CO LTD (JP)
International Classes:
G06N20/00; G06N3/12; G10L25/39
Foreign References:
JP2011255106A2011-12-22
KR101090892B12011-12-13
US20070168306A12007-07-19
JP2003099084A2003-04-04
JP2017067929A2017-04-06
Other References:
ODAWARA, MASARU ET AL.: "Feature selection based on genetic algorithm with mutual information filtering", FIT2008 FORUM ON INFORMATION TECHNOLOGY 2008, 20 August 2008 (2008-08-20), pages 473 - 474
Attorney, Agent or Firm:
Eikoh Patent Firm, P.C. (JP)
Download PDF: