A Comparative Study on Effect of Temporal Phase for Speaker Verification

Nongrum, Doreen; Pyrtuh, Fidalizia

doi:10.1007/978-981-19-0105-8_56

Cited by 3 publications

(1 citation statement)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such features can improve the performance by adapting the phase to tasks. The phase-based features have been used for speaker recognition and spoofed speech discrimination from genuine speech in recent past [12,13].…”

Section: Introductionmentioning

confidence: 99%

Phase characteristics of vocal tract filter can distinguish speakers

Okada,

Ito

2023

Front. Appl. Math. Stat.

View full text Add to dashboard Cite

IntroductionSpeaker recognition has been performed by considering individual variations in the power spectrograms of speech, which reflect the resonance phenomena in the speaker's vocal tract filter. In recent years, phase-based features have been used for speaker recognition. However, the phase-based features are not in a raw form of the phase but are crafted by humans, suggesting that the role of the raw phase is less interpretable. This study used phase spectrograms, which are calculated by subtracting the phase in the time-frequency domain of the electroglottograph signal from that of speech. The phase spectrograms represent the non-modified phase characteristics of the vocal tract filter.MethodsThe phase spectrograms were obtained from five Japanese participants. Phase spectrograms corresponding to vowels, called phase spectra, were then extracted and circular-averaged for each vowel. The speakers were determined based on the degree of similarity of the averaged spectra.ResultsThe accuracy of discriminating speakers using the averaged phase spectra was observed to be high although speakers were discriminated using only phase information without power. In particular, the averaged phase spectra showed different shapes for different speakers, resulting in the similarity between the different speaker spectrum pairs being lower. Therefore, the speakers were distinguished by using phase spectra.DiscussionThis predominance of phase spectra suggested that the phase characteristics of the vocal tract filter reflect the individuality of speakers.

show abstract

Section: Introductionmentioning

confidence: 99%

Phase characteristics of vocal tract filter can distinguish speakers

Okada,

Ito

2023

Front. Appl. Math. Stat.

View full text Add to dashboard Cite

show abstract

Feature Extraction Analysis in a Speaker Identification System

Goel,

Sharma,

Gambhir

et al. 2024

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

Analysis of Deep Generative Model Impact on Feature Extraction and Dimension Reduction for Short Utterance Text-Independent Speaker Verification

Farhadipour,

Veisi

2024

Circuits Syst Signal Process

View full text Add to dashboard Cite

Speaker verification is a biometric-based method for individual authentication. However, there are still several challenging problems in achieving high performance in short utterance text-independent conditions, maybe for weak speaker-specific features. Recently, deep learning algorithms have been used extensively in speech processing. This manuscript uses a deep belief network (DBN) as a deep generative method for feature extraction in speaker verification systems. This study aims to show the impact of using the proposed method in various challenging issues, including short utterances, text independence, language variation, and large-scale speaker verification. The proposed DBN uses MFCC as input and tries to extract more efficient features. This new representation of speaker information is evaluated in two popular speaker verification systems: GMM-UBM and i-vector-PLDA methods. The results show that, for the i-vector-PLDA system, the proposed feature decreases the EER considerably from 15.24 to 10.97%. In another experiment, DBN is used to reduce feature dimension and achieves significant results in decreasing computational time and increasing system response speed. In a case study, all the evaluations are performed for 1270 speakers of the NIST SRE2008 dataset. We show deep belief networks can be used in state-of-the-art acoustic modeling methods and more challenging datasets.

show abstract

A Comparative Study on Effect of Temporal Phase for Speaker Verification

Cited by 3 publications

References 29 publications

Phase characteristics of vocal tract filter can distinguish speakers

Phase characteristics of vocal tract filter can distinguish speakers

Feature Extraction Analysis in a Speaker Identification System

Analysis of Deep Generative Model Impact on Feature Extraction and Dimension Reduction for Short Utterance Text-Independent Speaker Verification

Contact Info

Product

Resources

About