Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
SPEECH ENHANCEMENT METHOD COMBINED WITH AI MODEL, SYSTEM, ELECTRONIC DEVICE AND MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/179424
Kind Code:
A1
Abstract:
A speech enhancement method combined with an AI model, a system, an electronic device and a medium. The speech enhancement method combined with an AI model comprises the following steps: acquiring a speech existence probability according to a target AI model (S11); obtaining noise power according to the speech existence probability (S12); obtaining a priori signal-to-noise ratio according to the noise power (S13); and obtaining a speech enhancement gain according to the prior signal-to-noise ratio (S14). The present invention can improve the intelligibility of a speech in a complex noisy scenario, and can also improve the performance of voice trigger detection and speech detection functions.

Inventors:
KANG LI (CN)
YE SHUNZHOU (CN)
LU CHENG (CN)
BA LIFANG (CN)
Application Number:
PCT/CN2020/088399
Publication Date:
September 16, 2021
Filing Date:
April 30, 2020
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
UNISOC CHONGQING TECH CO LTD (CN)
International Classes:
G10L21/0216; G10L25/21
Foreign References:
CN110739005A2020-01-31
CN110390950A2019-10-29
CN108831499A2018-11-16
CN110634500A2019-12-31
CN109979478A2019-07-05
US6782362B12004-08-24
Attorney, Agent or Firm:
SHANGHAI BESHINING LAW OFFICE (CN)
Download PDF: