In-Set/Out-of-Set Speaker Recognition Under Sparse Enrollment

Prakash, Vinod; Hansen, John H. L.

doi:10.1109/tasl.2007.902058

Cited by 22 publications

(29 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The concept of GMM-UBM (Reynolds et al 2000) is widely used for speaker recognition where the availability of training data is sparse (Angkititrakul & Hansen 2007, Prakash & Hansen 2007. In case of GMM-UBM system, speech data collected from large number of speakers is pooled and the UBM is trained.…”

Section: Speaker Modelling Using Gmm-ubmmentioning

confidence: 99%

“…The UBM training can be done in two ways: (i) Speech data pooled from the other database, not used for the speaker recognition study, provided speech data is collected from the same environment. (ii) Same speech data for both UBM training and evaluation, provided the speakers set used for recognition is not included in UBM training (Angkititrakul & Hansen 2007, Prakash & Hansen 2007, Reynolds et al 2000. We conducted the study using the YOHO database for both UBM training and evaluation.…”

Section: Speaker Modelling Using Gmm-ubmmentioning

confidence: 99%

“…The success of GMM is due to the availability of sufficient data for speaker modelling (Angkititrakul & Hansen 2007). Recently, some attempts have been made to recognize the speakers under limited data condition using the concept of Gaussian Mixture Model-Universal Background Model (GMM-UBM) (Angkititrakul & Hansen 2007) (Prakash & Hansen 2007). In another attempt the authors have proposed that by selecting those feature vectors which provide good speaker discrimination, it is possible to identify speakers under limited data (Kwon & Narayanan 2007).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

An experimental comparison of modelling techniques for speaker recognition under limited data condition

Jayanna

Prasanna

2009

Sadhana

View full text Add to dashboard Cite

Section: Speaker Modelling Using Gmm-ubmmentioning

confidence: 99%

Section: Speaker Modelling Using Gmm-ubmmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An experimental comparison of modelling techniques for speaker recognition under limited data condition

Jayanna

Prasanna

2009

Sadhana

View full text Add to dashboard Cite

“…The Gaussian mixture model (GMM) employing universal background model (UBM) with MAP speaker adaptation is the dominant approach in text-independent speaker recognition (Reynolds 2000;Prakash & Hansen 2007). GMM basically models the feature vectors of the speaker as Gaussian densities.…”

Section: Speaker Modellingmentioning

confidence: 99%

Application of combined temporal and spectral processing methods for speaker recognition under noisy, reverberant or multi-speaker environments

Krishnamoorthy

Prasanna

2009

Sadhana

View full text Add to dashboard Cite

This paper presents an experimental evaluation of the combined temporal and spectral processing methods for speaker recognition task under noise, reverberation or multi-speaker environments. Automatic speaker recognition system gives good performance in controlled environments. Speech recorded in real environments by distant microphones is degraded by factors like background noise, reverberation and interfering speakers. This degradation strongly affects the performance of the speaker recognition system. Combined temporal and spectral processing (TSP) methods proposed in our earlier study are used for pre-processing to improve the speaker-specific features and hence the speaker recognition performance. Different types of degradation like background noise, reverberation and interfering speaker are considered for evaluation. The evaluation is carried out for the individual temporal processing, spectral processing and the combined TSP method. The experimental results show that the combined TSP methods give relatively higher recognition performance compared to either temporal or spectral processing alone.

show abstract

“…In the case of limited training data, it may not be feasible to reliably estimate the parameters of such a statistical model directly. As a solution, parameters can be adapted from a universal background model (UBM) [25], [26]. Recently, speaker identification methods based on similarity rather than probability, such as fuzzy vector quantization [27], [28], have been developed as an alternative solution to the problem of limited training data.…”

mentioning

confidence: 99%