Further optimisations of constant Q cepstral processing for integrated utterance and text-dependent speaker verification

Delgado, Héctor; Todisco, Massimiliano; Sahidullah, Md; Sarkar, Achintya Kumar; Evans, Nicholas; Kinnunen, Tomi

doi:10.1109/slt.2016.7846262

Cited by 32 publications

(26 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A standard Gaussian mixture model with universal background model (GMM-UBM) ASV system was used for ranking [13]. It uses a Mel-frequency cepstral coefficient (MFCC) front-end and a 512-component UBM trained using RSR2015 [14] and TIMIT 7 databases.…”

Section: Replay Configurationsmentioning

confidence: 99%

ASVspoof 2017 Version 2.0: meta-data analysis and baseline enhancements

Delgado

Todisco²,

Sahidullah

et al. 2018

The Speaker and Language Recognition Workshop (Odyssey 2018)

Self Cite

123

110

View full text Add to dashboard Cite

The now-acknowledged vulnerabilities of automatic speaker verification (ASV) technology to spoofing attacks have spawned interests to develop so-called spoofing countermeasures. By providing common databases, protocols and metrics for their assessment, the ASVspoof initiative was born to spearhead research in this area. The first competitive ASVspoof challenge held in 2015 focused on the assessment of countermeasures to protect ASV technology from voice conversion and speech synthesis spoofing attacks. The second challenge switched focus to the consideration of replay spoofing attacks and countermeasures. This paper describes Version 2.0 of the ASVspoof 2017 database which was released to correct data anomalies detected post-evaluation. The paper contains as-yet unpublished meta-data which describes recording and playback devices and acoustic environments. These support the analysis of replay detection performance and limits. Also described are new results for the official ASVspoof baseline system which is based upon a constant Q cesptral coefficient frontend and a Gaussian mixture model backend. Reported are enhancements to the baseline system in the form of log-energy coefficients and cepstral mean and variance normalisation in addition to an alternative i-vector backend. The best results correspond to a 48% relative reduction in equal error rate when compared to the original baseline system.

show abstract

Section: Replay Configurationsmentioning

confidence: 99%

ASVspoof 2017 Version 2.0: meta-data analysis and baseline enhancements

Delgado

Todisco²,

Sahidullah

et al. 2018

The Speaker and Language Recognition Workshop (Odyssey 2018)

Self Cite

123

110

View full text Add to dashboard Cite

show abstract

“…We chose a lightweight GMM-UBM-based system to conduct rapid parameter experimentation with computationally heavy 2DAR models. As demonstrated in [21], GMM-UBM provides a competitive accuracy on the RedDots data consisting of short utterances. Figure 4 presents the results for the speaker verification experiments on RedDots corpus in terms of EER.…”

Section: Features and Classifiermentioning

confidence: 92%

Time-Varying Autoregressions for Speaker Verification in Reverberant Conditions

Vestman¹,

Gowda²,

Sahidullah³

et al. 2017

Interspeech 2017

Self Cite

View full text Add to dashboard Cite

In poor room acoustics conditions, speech signals received by a microphone might become corrupted by the signals' delayed versions that are reflected from the room surfaces (e.g. wall, floor). This phenomenon, reverberation, drops the accuracy of automatic speaker verification systems by causing mismatch between the training and testing. Since reverberation causes temporal smearing to the signal, one way to tackle its effects is to study robust feature extraction, particularly based on long-time temporal feature extraction. This approach has been adopted previously in the form of 2-dimensional autoregressive (2DAR) feature extraction scheme by using frequency domain linear prediction (FDLP). In 2DAR, FDLP processing is followed by time domain linear prediction (TDLP). In the current study, we propose modifying the latter part of the 2DAR feature extraction scheme by replacing TDLP with time-varying linear prediction (TVLP) to add an extra layer of temporal processing. Our speaker verification experiments using the proposed features with the text-dependent RedDots corpus show small but consistent improvements in clean and reverberant conditions (up to 6.5%) over the 2DAR features and large improvements over the MFCC features in reverberant conditions (up to 46.5%).

show abstract

“…They include (i) a standard Mel-frequency cepstral coefficients (MFCC) [14] frontend and (ii) an infinite impulse response -constant Q, Melfrequency cepstral coefficients (ICMC) [15] frontend. The latter has been applied successfully to tasks including speaker recognition, utterance verification [15] and speaker diarization [9,16]. These features are similar to MFCC, but they replace the short-time Fourier transform by an infinite-impulse re- sponse, constant Q transform (IIR-CQT) [17].…”

Section: Feature Extractionmentioning

confidence: 99%

ODESSA at Albayzin Speaker Diarization Challenge 2018

Patino¹,

Delgado

Yin

et al. 2018

IberSPEECH 2018

Self Cite

View full text Add to dashboard Cite

This paper describes the ODESSA submissions to the Albayzin Speaker Diarization Challenge 2018. The challenge addresses the diarization of TV shows. This work explores three different techniques to represent speech segments, namely binary key, x-vector and triplet-loss based embeddings. While training-free methods such as the binary key technique can be applied easily to a scenario where training data is limited, the training of robust neural-embedding extractors is considerably more challenging. However, when training data is plentiful (open-set condition), neural embeddings provide more robust segmentations, giving speaker representations which lead to better diarization performance. The paper also reports our efforts to improve speaker diarization performance through system combination. For systems with a common temporal resolution, fusion is performed at segment level during clustering. When the systems under fusion produce segmentations with an arbitrary resolution, they are combined at diarization hypothesis level. Both approaches to fusion are shown to improve diarization performance.

show abstract

Further optimisations of constant Q cepstral processing for integrated utterance and text-dependent speaker verification

Cited by 32 publications

References 22 publications

ASVspoof 2017 Version 2.0: meta-data analysis and baseline enhancements

ASVspoof 2017 Version 2.0: meta-data analysis and baseline enhancements

Time-Varying Autoregressions for Speaker Verification in Reverberant Conditions

ODESSA at Albayzin Speaker Diarization Challenge 2018

Contact Info

Product

Resources

About