Title:
VOICE INFORMATION PROCESSING DEVICE AND METHOD
Document Type and Number:
Japanese Patent JP2020038315
Kind Code:
A
Abstract:
To perform high performance sound source separation even when sparsity of a sound source in time-frequency domain does not hold.SOLUTION: The voice information processing device comprises an input device, an output device, and a processing unit. The processing unit includes a time-frequency domain conversion unit to convert a voice signal input from the input device into a time-frequency domain, a feature amount calculation unit to calculate a feature amount from the time-frequency domain, and a first time-frequency mask estimation unit to estimate a first time-frequency mask for separating sound of a specific direction from the feature amount.SELECTED DRAWING: Figure 1
Inventors:
IKESHITA RINTARO
FUJIOKA TAKUYA
HORIGUCHI SHOTA
FUJIOKA TAKUYA
HORIGUCHI SHOTA
Application Number:
JP2018165875A
Publication Date:
March 12, 2020
Filing Date:
September 05, 2018
Export Citation:
Assignee:
HITACHI LTD
International Classes:
G10L21/0272; G10L21/028; G10L25/51; H04R3/00
Attorney, Agent or Firm:
Aoritsu patent business corporation