Title:
METHOD, APPARATUS, AND DEVICE FOR TRAINING MULTIMODE VOICE RECOGNITION MODEL, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/196802
Kind Code:
A1
Abstract:
A method, apparatus, and device for training a multimode voice recognition model, and a storage medium. During the training of a multimode voice recognition model, training data comprises pure audio signals and a data set used for generating corresponding image features on the basis of the pure audio signals. A training data set during the training of the multimode voice recognition model is enriched, so that the generalization capability of a multimode voice processing method is improved, and the reliability of a multimode voice recognition model is improved.
Inventors:
JING ZIJUN (CN)
PAN JIA (CN)
WU HUAXIN (CN)
PAN JIA (CN)
WU HUAXIN (CN)
Application Number:
PCT/CN2020/142166
Publication Date:
October 07, 2021
Filing Date:
December 31, 2020
Export Citation:
Assignee:
IFLYTEK CO LTD (CN)
International Classes:
G10L15/02; G06K9/62; G10L15/25
Foreign References:
CN111462733A | 2020-07-28 | |||
CN108389573A | 2018-08-10 | |||
CN110544479A | 2019-12-06 | |||
CN105022470A | 2015-11-04 | |||
CN102023703A | 2011-04-20 | |||
US20170278517A1 | 2017-09-28 | |||
US20190371334A1 | 2019-12-05 |
Attorney, Agent or Firm:
UNITALEN ATTORNEYS AT LAW (CN)
Download PDF: