Title:
METHOD AND APPARATUS FOR SPEECH SIGNAL ESTIMATION USING ATTENTION MECHANISM
Document Type and Number:
WIPO Patent Application WO/2022/158914
Kind Code:
A1
Abstract:
A multi-channel-based noise and echo signal integrated cancellation apparatus using a deep neural network, according to an embodiment, may comprise: a plurality of microphone encoders that receive inputs of a plurality of microphone input signals including an echo signal, a noise signal, and a speech signal of an utterer, convert the plurality of microphone input signals respectively into a plurality of pieces of conversion information, and output same; a channel conversion unit that compresses the plurality of pieces of conversion information and converts them into first input information having the size of a single channel, and outputs same; a far-end signal encoder that receives an input of a far-end signal, converts the far-end signal into second input information, and outputs same; an attention unit that outputs weight information by applying an attention mechanism to the first input information and the second input information; a pre-trained first artificial neural network that uses, as input information, third input information which is information that is the sum of the weight information and the second input information, and uses, as output information, first output information including mask information for estimating the speech signal from the second input information; and a speech signal estimation unit that outputs an estimation speech signal estimated by a speech signal unit, on the basis of the first output information and the second input information.
More Like This:
Inventors:
CHANG JOON HYUK (KR)
PARK SONG KYU (KR)
PARK SONG KYU (KR)
Application Number:
PCT/KR2022/001166
Publication Date:
July 28, 2022
Filing Date:
January 21, 2022
Export Citation:
Assignee:
IUCF HYU (KR)
International Classes:
G10L21/0272; G06N3/08; G10L21/0208; G10L25/30
Foreign References:
| KR20200115107A | 2020-10-07 | |||
| US20180040333A1 | 2018-02-08 |
Other References:
FAZEL AMIN; EL-KHAMY MOSTAFA; LEE JUNGWON: "CAD-AEC: Context-Aware Deep Acoustic Echo Cancellation", ICASSP 2020 - 2020 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 4 May 2020 (2020-05-04), pages 6919 - 6923, XP033793171, DOI: 10.1109/ICASSP40776.2020.9053508
GIRI RITWIK; ISIK UMUT; KRISHNASWAMY ARVINDH: "Attention Wave-U-Net for Speech Enhancement", 2019 IEEE WORKSHOP ON APPLICATIONS OF SIGNAL PROCESSING TO AUDIO AND ACOUSTICS (WASPAA), 20 October 2019 (2019-10-20), pages 249 - 253, XP033677307, DOI: 10.1109/WASPAA.2019.8937186
KIM JUNG-HEE, CHANG JOON-HYUK: "Attention Wave-U-Net for Acoustic Echo Cancellation", INTERSPEECH 2020, 1 October 2020 (2020-10-01) - 29 October 2020 (2020-10-29), ISCA, pages 3969 - 3973, XP055952817, DOI: 10.21437/Interspeech.2020-3200
BARMPOUTIS PANAGIOTIS, PAPAIOANNOU PERIKLIS, DIMITROPOULOS KOSMAS, GRAMMALIDIS NIKOS: "A Review on Early Forest Fire Detection Systems Using Optical Remote Sensing", SENSORS, vol. 20, no. 22, 11 November 2020 (2020-11-11), CH , pages 1 - 26, XP055946762, ISSN: 1424-8220, DOI: 10.3390/s20226442
GIRI RITWIK; ISIK UMUT; KRISHNASWAMY ARVINDH: "Attention Wave-U-Net for Speech Enhancement", 2019 IEEE WORKSHOP ON APPLICATIONS OF SIGNAL PROCESSING TO AUDIO AND ACOUSTICS (WASPAA), 20 October 2019 (2019-10-20), pages 249 - 253, XP033677307, DOI: 10.1109/WASPAA.2019.8937186
KIM JUNG-HEE, CHANG JOON-HYUK: "Attention Wave-U-Net for Acoustic Echo Cancellation", INTERSPEECH 2020, 1 October 2020 (2020-10-01) - 29 October 2020 (2020-10-29), ISCA, pages 3969 - 3973, XP055952817, DOI: 10.21437/Interspeech.2020-3200
BARMPOUTIS PANAGIOTIS, PAPAIOANNOU PERIKLIS, DIMITROPOULOS KOSMAS, GRAMMALIDIS NIKOS: "A Review on Early Forest Fire Detection Systems Using Optical Remote Sensing", SENSORS, vol. 20, no. 22, 11 November 2020 (2020-11-11), CH , pages 1 - 26, XP055946762, ISSN: 1424-8220, DOI: 10.3390/s20226442
Attorney, Agent or Firm:
HAEUM PATENT & LAW FIRM (KR)
Download PDF: