Title:
逆離散コサイン変換のケプストラム係数の動的分割に基づく音声特徴抽出アルゴリズム
Document Type and Number:
Japanese Patent JP6783001
Kind Code:
B2
Abstract:
The invention discloses a sound characteristic extraction algorithm based on dynamic partitioned inverse discrete cosine transform cepstral coefficients. The method comprises the following steps: S1,carrying out pre-emphasis, framing and windowing preprocessing on a sound signal; S2, carrying out transform mode processing on the preprocessed sound signal from a time domain to a frequency domain;and S3, calculating the similarity of inverse discrete cosine transform cepstral coefficients obtained in the step S2 by a cluster analysis algorithm, sequentially combining two adjacent types with the maximum similarity, and carrying out iteration on the process till 24 types are clustered, thereby obtaining dynamic partitioned inverse discrete cosine transform cepstral coefficients, that is, sound characteristics. By adopting the algorithm, the defect that in the prior art, dynamic sound characteristics are not sufficiently utilized for frequency domain transform is avoided, the algorithm has wide adaptability, and in addition, high recognition precision can be achieved in speaker recognition.
Inventors:
Left
Ma Hee
Lee Iron Mountain
Kabei Super
Liu Kimi
艾 佳▲チー▼
Xiao Yang
In renkai
Ma Hee
Lee Iron Mountain
Kabei Super
Liu Kimi
艾 佳▲チー▼
Xiao Yang
In renkai
Application Number:
JP2019186806A
Publication Date:
November 11, 2020
Filing Date:
October 10, 2019
Export Citation:
Assignee:
Dalian Maritime University
International Classes:
G10L25/24
Domestic Patent References:
| JP2004093671A | ||||
| JP2002183668A | ||||
| JP7160287A | ||||
| JP7098598A |
Foreign References:
| US20180005628 |
Other References:
R. Muralishankar, A. G. Ramakrishnan,DCT based pseudo complex cepstrum,IEEE International Conference on Acoustics, Speech, and Signal Processing,2002年 5月,p. I-521~I-524
MATLAB Note/音声の分析 - Miyazawa's Pukiwiki,[online],2018年12月12日,[2019年11月29日検索],インターネット,URL,https://web.archive.org/web/20181212203646/http://speechresearch.fiw-web.net/73.html
林史尊,外3名,GPUを用いたCanopyクラスタリングの高速化,日本データベース学会和文論文誌,2015年 2月23日,第13-J巻,第2号,p.13-18
MATLAB Note/音声の分析 - Miyazawa's Pukiwiki,[online],2018年12月12日,[2019年11月29日検索],インターネット,URL,https://web.archive.org/web/20181212203646/http://speechresearch.fiw-web.net/73.html
林史尊,外3名,GPUを用いたCanopyクラスタリングの高速化,日本データベース学会和文論文誌,2015年 2月23日,第13-J巻,第2号,p.13-18
Attorney, Agent or Firm:
Kimura Mitsuru
Taiji Morikawa
Kei Sakurada
Mie Hideki
Taiji Morikawa
Kei Sakurada
Mie Hideki