Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
CONTINUOUS NOISE TRACKING-BASED TARGET SPEECH SIGNAL ENHANCEMENT METHOD AND SYSTEM, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2020/177374
Kind Code:
A1
Abstract:
A continuous noise tracking-based target speech signal enhancement method and system, and a storage medium; the method comprises: step 1: receiving a speech signal that has noise, performing framing windowing processing on the speech signal that has noise, and using short-time Fourier transform to obtain a time-frequency domain relationship; step 2: estimating a noise power spectrum; step 3: estimating a speech power spectrum; step 4: estimating a speech signal by means of a speech estimator; and step 5: performing inverse Fourier transform, windowing and using an overlap adding technique to achieve speech recovery. The present method effectively separates the target speech signal, and reduces the residual amount of noise in the speech signal so that the quality of the target signal is improved. The present invention plays an important role in applications such as automatic speech recognition, speaker recognition, human-machine dialogue interfaces, and hearing aids.

Inventors:
ZHANG QIQUAN (CN)
WANG MINGJIANG (CN)
LU YUN (CN)
HAN YUFEI (CN)
ZHANG LU (CN)
SUN FENGJIAO (CN)
Application Number:
PCT/CN2019/117078
Publication Date:
September 10, 2020
Filing Date:
November 11, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
HARBIN INST TECHNOLOGY SHENZHEN (CN)
International Classes:
G10L21/02; G10L21/0272; G10L25/03
Foreign References:
CN109817234A2019-05-28
CN109308904A2019-02-05
CN106971740A2017-07-21
CN106448692A2017-02-22
US6289309B12001-09-11
CN108074582A2018-05-25
CN108735225A2018-11-02
Attorney, Agent or Firm:
SZ KINDWALF INTELLECTUAL PROPERTY FIRM (CN)
Download PDF: