Title:
TRAINING SYSTEM AND METHOD FOR ACOUSTIC MODEL
Document Type and Number:
WIPO Patent Application WO/2024/075633
Kind Code:
A1
Abstract:
A training system for an acoustic model can make various trainings easily implementable by making it possible to select data to be used for training the acoustic model from a plurality of pieces of training data, and includes a first device that is used by a first user and connectable to a network, and a server that is connectable to the network. The first device uploads a plurality of sound waveforms to the server under the control of the first user, selects, from the plurality of sound waveforms that are already uploaded or to be uploaded, one or more sound waveforms as a first waveform set, and transmits, to the server, a first execution instruction of a first training job for an acoustic model for generating an acoustic feature amount. The server starts the execution of the first training job using the selected first waveform set on the basis of the first execution instruction from the first device, and provides, to the first device, a trained acoustic model trained by the first training job.
Inventors:
DAIDO RYUNOSUKE (JP)
SAINO KEIJIRO (JP)
NISHIMURA MASANARI (JP)
SHIMIZU MASAHIRO (JP)
SAINO KEIJIRO (JP)
NISHIMURA MASANARI (JP)
SHIMIZU MASAHIRO (JP)
Application Number:
PCT/JP2023/035432
Publication Date:
April 11, 2024
Filing Date:
September 28, 2023
Export Citation:
Assignee:
YAMAHA CORP (JP)
International Classes:
G10L13/06; G10H1/00; G10L13/00; G10K15/04
Attorney, Agent or Firm:
TAKAHASHI, HAYASHI AND PARTNER PATENT ATTORNEYS, INC. (JP)
Download PDF:
Previous Patent: DEVICE FOR PRODUCING INFLATION FILM
Next Patent: DISPLAY METHOD RELATED TO CHARACTERISTIC DISTRIBUTION OF SOUND WAVEFORM
Next Patent: DISPLAY METHOD RELATED TO CHARACTERISTIC DISTRIBUTION OF SOUND WAVEFORM