Studies on inter-speaker variability in speech and its application in automatic speech recognition

Umesh, S.

doi:10.1007/s12046-011-0049-x

Cited by 6 publications

(3 citation statements)

References 57 publications

(55 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most efficient and widely used acoustic parameters are Mel Frequency Cepstral Coefficients (MFCC) [2], [3]. The computation of these coefficients involves multiple steps, and the selection of these parameters is based on their ability for interpolation, robustness to noise, and adaptation to inter and intra-speaker variability [4] [5].…”

Section: Introductionmentioning

confidence: 99%

WITHDRAWN: An Experiment of Moroccan Dialect speech recognition in noisy environments using PocketSphinx

OUISAADANE,

SAFI,

FRIKEL

2023

Preprint

View full text Add to dashboard Cite

In this study, we introduce an experimental framework for Moroccan Dialect speech recognition under various additive noise conditions using the open-source tool PocketSphinx. We curated a corpus comprising the ten most commonly used greetings in the Moroccan dialect, extracted from telephone conversations. This corpus was recorded with 60 speakers (30 males and 30 females). Each speaker articulated each expression three times in natural and noisy conditions. Feature extraction utilized Mel Scale Cepstral Coefficients (MFCC), and acoustic modeling, based on monophony, was implemented using Hidden Markov Models (HMM). While automatic speech recognition systems demonstrate commendable performance in noise-free conditions, their efficacy noticeably diminishes in the presence of noise.

show abstract

Section: Introductionmentioning

confidence: 99%

WITHDRAWN: An Experiment of Moroccan Dialect speech recognition in noisy environments using PocketSphinx

OUISAADANE,

SAFI,

FRIKEL

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Emotion is one of the internal sources that most affects the performance of the speaker verification system. Emotion induce intra-speaker vocal variability [18], even more for stress condition [19].…”

Section: Related Workmentioning

confidence: 99%

Emotional Variability Analysis Based I-Vector for Speaker Verification in Under-Stress Conditions

2020

View full text Add to dashboard Cite

Emotional conditions cause changes in the speech production system. It produces the differences in the acoustical characteristics compared to neutral conditions. The presence of emotion makes the performance of a speaker verification system degrade. In this paper, we propose a speaker modeling that accommodates the presence of emotions on the speech segments by extracting a speaker representation compactly. The speaker model is estimated by following a similar procedure to the i-vector technique, but it considerate the emotional effect as the channel variability component. We named this method as the emotional variability analysis (EVA). EVA represents the emotion subspace separately to the speaker subspace, like the joint factor analysis (JFA) model. The effectiveness of the proposed system is evaluated by comparing it with the standard i-vector system in the speaker verification task of the Speech Under Simulated and Actual Stress (SUSAS) dataset with three different scoring methods. The evaluation focus in terms of the equal error rate (EER). In addition, we also conducted an ablation study for a more comprehensive analysis of the EVA-based i-vector. Based on experiment results, the proposed system outperformed the standard i-vector system and achieved state-of-the-art results in the verification task for the under-stressed speakers.

show abstract

“…An important factor that characterises speech signals is vocal tract length (VTL), which varies from person to person and is of importance to deal with in speech systems including speaker recognition and automatic speech recognition (ASR). For example, vocal tract length normalization (VTLN) [1], [2], [3] is widely used in ASR to reduce the effect of speaker variability due to the difference in VTL among speakers. In [4], VTL warped features are used for ASR as a data augmentation technique.…”

Section: Introductionmentioning

confidence: 99%

Vocal Tract Length Perturbation for Text-Dependent Speaker Verification with Autoregressive Prediction Coding

Sarkar,

Tan

2020

Preprint

View full text Add to dashboard Cite

In this letter, we propose a vocal tract length (VTL) perturbation method for text-dependent speaker verification (TD-SV), in which a set of TD-SV systems are trained, one for each VTL factor, and score-level fusion is applied to make a final decision. Next, we explore the bottleneck (BN) feature extracted by training deep neural networks with a self-supervised objective, autoregressive predictive coding (APC), for TD-SV and compare it with the well-studied speaker-discriminant BN feature. The proposed VTL method is then applied to APC and speakerdiscriminant BN features. In the end, we combine the VTL perturbation systems trained on MFCC and the two BN features in the score domain. Experiments are performed on the RedDots challenge 2016 database of TD-SV using short utterances with Gaussian mixture model-universal background model and ivector techniques. Results show the proposed methods significantly outperform the baselines.

show abstract

Studies on inter-speaker variability in speech and its application in automatic speech recognition

Cited by 6 publications

References 57 publications

WITHDRAWN: An Experiment of Moroccan Dialect speech recognition in noisy environments using PocketSphinx

WITHDRAWN: An Experiment of Moroccan Dialect speech recognition in noisy environments using PocketSphinx

Emotional Variability Analysis Based I-Vector for Speaker Verification in Under-Stress Conditions

Vocal Tract Length Perturbation for Text-Dependent Speaker Verification with Autoregressive Prediction Coding

Contact Info

Product

Resources

About