Improving anti-spoofing with octave spectrum and short-term spectral statistics information. (1st January 2020)
- Record Type:
- Journal Article
- Title:
- Improving anti-spoofing with octave spectrum and short-term spectral statistics information. (1st January 2020)
- Main Title:
- Improving anti-spoofing with octave spectrum and short-term spectral statistics information
- Authors:
- Yang, Jichen
Das, Rohan Kumar - Abstract:
- Abstract: The long-term window based features have been found to be effective for spoofing attack detection. One such important countermeasure is constant-Q cepstral coefficients (CQCC) that is derived from constant-Q transform. During its extraction, the octave power spectrum is converted to the linear power spectrum by performing uniform resampling. However, the information from the octave power spectrum is different from that carried by the linear power spectrum. We believe that the octave power spectrum can offer complementary information to the linear power spectrum for spoofing attack detection. In this regard, we propose to combine the coefficients generated using both linear and octave power spectrum. The combined feature is referred to as extended CQCC (eCQCC), which is hypothesized to have better discriminative information for detection of spoofing attacks. In addition, we use the short-term spectral statistics information (STSSI) along with eCQCC feature to form another novel feature representation referred to as eCQCC-STSSI to have improved anti-spoofing countermeasure. We perform the studies with the proposed features for both synthetic and playback attacks using ASVspoof 2015 and ASVspoof 2017 version 2.0 corpus, respectively. The studies reveal that eCQCC outperforms the conventional CQCC feature as well most of the known systems showing importance of octave spectrum information. Further, the hybrid feature eCQCC-STSSI improve the performance of eCQCC featureAbstract: The long-term window based features have been found to be effective for spoofing attack detection. One such important countermeasure is constant-Q cepstral coefficients (CQCC) that is derived from constant-Q transform. During its extraction, the octave power spectrum is converted to the linear power spectrum by performing uniform resampling. However, the information from the octave power spectrum is different from that carried by the linear power spectrum. We believe that the octave power spectrum can offer complementary information to the linear power spectrum for spoofing attack detection. In this regard, we propose to combine the coefficients generated using both linear and octave power spectrum. The combined feature is referred to as extended CQCC (eCQCC), which is hypothesized to have better discriminative information for detection of spoofing attacks. In addition, we use the short-term spectral statistics information (STSSI) along with eCQCC feature to form another novel feature representation referred to as eCQCC-STSSI to have improved anti-spoofing countermeasure. We perform the studies with the proposed features for both synthetic and playback attacks using ASVspoof 2015 and ASVspoof 2017 version 2.0 corpus, respectively. The studies reveal that eCQCC outperforms the conventional CQCC feature as well most of the known systems showing importance of octave spectrum information. Further, the hybrid feature eCQCC-STSSI improve the performance of eCQCC feature due to the STSSI information combined with it. … (more)
- Is Part Of:
- Applied acoustics. Volume 157(2020)
- Journal:
- Applied acoustics
- Issue:
- Volume 157(2020)
- Issue Display:
- Volume 157, Issue 2020 (2020)
- Year:
- 2020
- Volume:
- 157
- Issue:
- 2020
- Issue Sort Value:
- 2020-0157-2020-0000
- Page Start:
- Page End:
- Publication Date:
- 2020-01-01
- Subjects:
- CQT spectrum -- Anti-spoofing countermeasure -- Octave power spectrum -- Short-term spectral statistics
Acoustical engineering -- Periodicals
Periodicals
620.2 - Journal URLs:
- http://www.sciencedirect.com/science/journal/0003682X ↗
http://www.elsevier.com/journals ↗
http://www.elsevier.com/homepage/elecserv.htt ↗ - DOI:
- 10.1016/j.apacoust.2019.107017 ↗
- Languages:
- English
- ISSNs:
- 0003-682X
- Deposit Type:
- Legaldeposit
- View Content:
- Available online (eLD content is only available in our Reading Rooms) ↗
- Physical Locations:
- British Library DSC - 1571.400000
British Library DSC - BLDSS-3PM
British Library HMNTS - ELD Digital store - Ingest File:
- 16409.xml