CBC-Based Synthetic Speech Detection

2019 ◽  
Vol 11 (2) ◽  
pp. 63-74 ◽  
Author(s):  
Jichen Yang ◽  
Qianhua He ◽  
Yongjian Hu ◽  
Weiqiang Pan
Keyword(s):  
Power Spectrum ◽  
Time Domain ◽  
Extraction Method ◽  
Synthetic Speech ◽  
Equal Error Rate ◽  
Speech Detection ◽  
Feature Extraction Method ◽  
Block Transform ◽  
New Feature ◽  
The Time Domain

In previous studies of synthetic speech detection (SSD), the most widely used features are based on a linear power spectrum. Different from conventional methods, this article proposes a new feature extraction method for SSD from octave power spectrum which is obtained from constant-Q transform (CQT). By combining CQT, block transform (BT) and discrete cosine transform (DCT), a new feature is obtained, namely, constant-Q block coefficients (CBC). In which, CQT is used to transform speech from the time domain into the frequency domain, BT is used to segment octave power spectrum into many blocks and DCT is used to extract principal information of every block. The experimental results on ASVspoof 2015 corpus shows that CBC is superior to other front-ends features that have been benchmarked on ASVspoof 2015 evaluation set in terms of equal error rate (EER).

scholarly journals A new feature extraction method for signal classification applied to cord dorsum potential detection

2012 ◽  
Vol 9 (5) ◽  
pp. 056009 ◽  
Author(s):  
D Vidaurre ◽  
E E Rodríguez ◽  
C Bielza ◽  
P Larrañaga ◽  
P Rudomin
Keyword(s):  
Feature Extraction ◽  
Extraction Method ◽  
Signal Classification ◽  
Feature Extraction Method ◽  
New Feature

A new feature extraction method for odour classification

2011 ◽  
Vol 158 (1) ◽  
pp. 75-88 ◽  
Author(s):  
Bernd Ehret ◽  
Konstantin Safenreiter ◽  
Frank Lorenz ◽  
Joachim Biermann
Keyword(s):  
Feature Extraction ◽  
Extraction Method ◽  
Feature Extraction Method ◽  
New Feature

scholarly journals A Vibration Feature Extraction Method Based on Time-Domain Dimensional Parameters and Mahalanobis Distance

2021 ◽  
Vol 2021 ◽  
pp. 1-12
Author(s):  
Junjun Chen ◽  
Bing Xu ◽  
Xin Zhang
Keyword(s):  
Feature Extraction ◽  
Frequency Domain ◽  
Time Domain ◽  
Mahalanobis Distance ◽  
Extraction Method ◽  
Vibration Signal ◽  
Feature Extraction Method ◽  
Feature Parameters ◽  
Dimensional Parameters ◽  
Failure Types

To accurately describe the characteristics of a signal, the feature parameters in time domain and frequency domain are usually extracted for characterization. However, the total number of feature parameters in time domain and frequency domain exceeds twenty, and all of the feature parameters are used for feature extraction, which will result in a large amount of data processing. For the purpose of using fewer feature parameters to accurately reflect the characteristics of the vibration signal, a simple but effective vibration feature extraction method combining time-domain dimensional parameters (TDDP) and Mahalanobis distance (MD) is proposed, i.e., TDDP-MD. In this method, ten time-domain dimensional parameters are selected to extract fault features, and the distance evaluation technique based on Mahalanobis distance criterion function is also introduced to calculate the feature vector, which can be used to classify different failure types. Finally, the proposed method is applied to fault diagnosis of rolling element bearings, and experimental analysis results show that the proposed method can recognize different failure types accurately and effectively with only ten time-domain dimensional parameters and a small quantity of training samples.

Sign in / Sign up

Export Citation Format

Share Document