Title:
AUDIO SIGNAL DATABASE GENERATION DEVICE, AND AUDIO SIGNAL RETRIEVING DEVICE
Document Type and Number:
WIPO Patent Application WO/2020/241073
Kind Code:
A1
Abstract:
Provided is database generation technology capable of accurately and efficiently generating a database which can be used in text-based audio signal retrieval. The present invention includes: a latent variable generation unit that uses an audio signal encoder to generate, from an audio signal, a latent variable corresponding to the audio signal; a data generation unit that uses a natural language expression decoder to generate a natural language expression corresponding to the audio signal from a condition relating to an index for a natural language expression and the latent variable; and an audio signal database generation unit that generates, from the natural language expression corresponding to the audio signal and the audio signal, a record containing the natural language expression corresponding to the audio signal and the audio signal, and generates an audio signal database comprising the record.
Inventors:
KASHINO KUNIO (JP)
IKAWA SHOTA (JP)
IKAWA SHOTA (JP)
Application Number:
PCT/JP2020/015794
Publication Date:
December 03, 2020
Filing Date:
April 08, 2020
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
UNIV TOKYO (JP)
UNIV TOKYO (JP)
International Classes:
G06F16/683; G10L15/00; G10L15/10; G10L15/16
Foreign References:
| JP2006058567A | 2006-03-02 | |||
| JP2002366552A | 2002-12-20 |
Other References:
IKAWA , SHOTA ET AL.: "Acoustic signal retrieval based on latent features using onomatopoeia as a query", ACOUSTICAL SOCIETY OF JAPAN 2018 AUTUMN MEETING, 14 September 2018 (2018-09-14), pages 927 - 930
Attorney, Agent or Firm:
NAKAO, Naoki et al. (JP)
Download PDF: