Discriminative training of GMM for speaker identification

Alamo, Cesar Martín del; Gil, F.J. Caminero; Munilla, C. dela Torre; Gómez, Luis A. Hernández

doi:10.1109/icassp.1996.540297

Cited by 16 publications

(5 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous studies show that MCE criterion creates significant improvements in performance for speaker recognition [3]. This study further investigated whether conducting MCE criterion on HCRF pursues additional improvements.…”

Section: Minimum Classification Error-based Training On Hcrfmentioning

confidence: 95%

“…In [3] the baseline GMMs are further refined through discriminative training algorithms to obtain more accurate speaker models and accordingly the error rate is significantly improved from 32% to 7%. Lee [4] employed two Mandarin speech databases, MAT2000 [5] and TCC-300 [6], for the experiments on the speaker identification; and the results conclude that the discriminative training could reduce the required amount of enrollment speech.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Speaker identification using Hidden Conditional Random Field-based speaker models

Hong

2010

2010 International Conference on Machine Learning and Cybernetics

View full text Add to dashboard Cite

In this paper we make a study of applying Hidden Conditional Random Fields (HCRF) to establish speaker models. A novel training algorithm combining the discriminative training criterion with HCRF for speaker identification is proposed. This work also adopted discriminative training technique to train GMM, HMM, and HCRF speaker models respectively; and the performance of speaker identification by the three speaker models with different amounts of training speech for clean and noisy testing speech were investigated. The experimental results indicate that the HCRF model consistently achieved the lowest error rate among the three models regardless of the length of the test and training speech and presence of noise.

show abstract

Section: Minimum Classification Error-based Training On Hcrfmentioning

confidence: 95%

Section: Introductionmentioning

confidence: 99%

Speaker identification using Hidden Conditional Random Field-based speaker models

Hong

2010

2010 International Conference on Machine Learning and Cybernetics

View full text Add to dashboard Cite

show abstract

“…In [4], segmental Generalized Probabilistic Descent (GPD) algorithm has been used to estimate model parameters of a class considering the competing speakers. Minimum Classification Errors(MCE) approach for speaker verification is proposed in [5].…”

Section: Introductionmentioning

confidence: 99%

Speaker identification using utterances correspond to speaker-specific-text

Bharathi¹,

Vijayalakshmi²,

Nagarajan³

2011

IEEE Technology Students' Symposium

View full text Add to dashboard Cite

In speaker recognition tasks, the main reason for reduced accuracy is due to closely resembling speakers in the acoustic space. Conventional GMM-based modelling technique captures unique features along with common features among various classes. Further, it ignores knowledge of phonetic content of the speech. In order to increase the discriminative power of the classifier, the system must be able to use only the unique features of a given speaker with respect to his/her acoustically closely resembling speaker. This paper proposes a technique to reduce the confusion errors, by finding speaker-specific phonemes and formulate a text using the subset of phonemes that are unique, for speaker identification task. Experiments have been conducted on speaker identification task using speech data of 192 female speakers from TIMIT corpus.The performance of the proposed system is compared with that of a conventional GMM-based technique and a significant improvement is noted.

show abstract

“…Gaussian Mixture Model (GMM) has been widely used in speaker identification and shows the good performance [10][11][12][13]. A Gaussian mixture density is a weighted sum of M component densities, as depicted in Fig.…”

Section: Matching Algorithmmentioning

confidence: 99%

Robust Speaker Identification System Based on Wavelet Transform and Gaussian Mixture Model

Chen

Hsieh

Lai

2005

Natural Language Processing – IJCNLP 2004

View full text Add to dashboard Cite

This paper presents an effective and robust method for extracting features for speech processing. Based on the time-frequency multiresolution property of wavelet transform, the input speech signal is decomposed into various frequency channels. For capturing the characteristics of the vocal track and vocal codes, the traditional linear predictive cepstral coefficients (LPCC) of the approximation channel, and the entropy of the detail channel for each decomposition process are calculated. In addition, a hard thresholding technique for each lower resolution is applied to remove interference from noise. Experimental results show that using this mechanism not only effectively reduces the influence of noise, but also improves recognition. Finally, the proposed feature extraction algorithm is evaluated on the MAT telephone speech database for text-independent speaker identification using the Gaussian Mixture Model (GMM) identifier. Some popular existing methods are also evaluated for comparison in this paper. The results show that the proposed method of feature extraction is more effective and robust than other methods. In addition, the performance of our method is very satisfactory even at low SNR.

show abstract

Discriminative training of GMM for speaker identification

Cited by 16 publications

References 5 publications

Speaker identification using Hidden Conditional Random Field-based speaker models

Speaker identification using Hidden Conditional Random Field-based speaker models

Speaker identification using utterances correspond to speaker-specific-text

Robust Speaker Identification System Based on Wavelet Transform and Gaussian Mixture Model

Contact Info

Product

Resources

About