• Proceedings of the KSPS conference
• /
• 2002.11a
• /
• pp.201-204
• /
• 2002

As a preliminary study for improving recognition performance of the connected digit telephone speech, we investigate feature parameters as well as channel compensation methods of telephone speech. The CMN and RTCN are examined for telephone channel compensation, and the MFCC, DWFBA, SSC and their delta-features are examined as feature parameters. Recognition experiments with database we collected show that in feature level DWFBA is better than MFCC and for channel compensation RTCN is better than CMN. The DWFBA+Delta_ Mel-SSC feature shows the highest recognition rate.

• MALSORI
• /
• no.57
• /
• pp.113-122
• /
• 2006

Recently, many studies have been done for speech recognition in noisy environments. Particularly, the Aurora DB has been built as the common database for comparing the various feature extraction schemes. However, in general, the recognition models as well as the features have to be modified for effective noisy speech recognition. As the structure of the HTK is very complex, it is not easy to modify, the recognition engine. In this paper, we implemented a baseline recognizer based on the segmental K-means algorithm whose performance is comparable to the HTK in spite of the simplicity in its implementation.

• ETRI Journal
• /
• v.28 no.4
• /
• pp.502-505
• /
• 2006

A new class-based histogram equalization method is proposed for robust speech recognition. The proposed method aims at not only compensating the acoustic mismatch between training and test environments, but also at reducing the discrepancy between the phonetic distributions of training and test speech data. The algorithm utilizes multiple class-specific reference and test cumulative distribution functions, classifies the noisy test features into their corresponding classes, and equalizes the features by using their corresponding class-specific reference and test distributions. Experiments on the Aurora 2 database proved the effectiveness of the proposed method by reducing relative errors by 18.74%, 17.52%, and 23.45% over the conventional histogram equalization method and by 59.43%, 66.00%, and 50.50% over mel-cepstral-based features for test sets A, B, and C, respectively.

• Proceedings of the KSPS conference
• /
• 2003.10a
• /
• pp.145-148
• /
• 2003

To improve the recognition performance of Korean connected digit telephone speech, in this paper, both Aurora feature extraction method that employs noise reduction 2-state Wiener filter and DWFBA method are investigated and used. CMN and MRTCN are applied to static features for channel compensation. Telephone digit speech database released by SITEC is used for recognition experiments with HTK system. Experimental results has shown that Aurora feature is slightly better than MFCC and DWFBA without channel compensation. And when channel compensation is included, Aurora feature is slightly better than DWFBA with MRTCN.

• Proceedings of the KSPS conference
• /
• 2006.05a
• /
• pp.135-138
• /
• 2006

The communication method between human and robot is one of the important parts for a human-robot interaction. And speech is easy and intuitive communication method for human-being. By using speech as a communication method for robot, we can use robot as familiar way. In this paper, we developed TTS for human-robot interaction. Synthesis algorithms were modified for an efficient utilization of restricted resource in robot. And synthesis database were reconstructed for an efficiency. As a result, we could reduce the computation time with slight degradation of the speech quality.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.3C
• /
• pp.257-267
• /
• 2009

This paper presents a training method for neural networks and the employment of MSE (mean scare error) values as the basis of a decision regarding the identity claim of a speaker in a recurrent neural networks based speaker verification system. Recurrent neural networks (RNNs) are employed to capture temporally dynamic characteristics of speech signal. In the process of supervised learning for RNNs, target outputs are automatically generated and the generated target outputs are made to represent the temporal variation of input speech sounds. To increase the capability of discriminating between the true speaker and an impostor, a discriminative training method for RNNs is presented. This paper shows the use and the effectiveness of the MSE value, which is obtained from the Euclidean distance between the target outputs and the outputs of networks for test speech sounds of a speaker, as the basis of speaker verification. In terms of equal error rates, results of experiments, which have been performed using the Korean speech database, show that the proposed speaker verification system exhibits better performance than a conventional hidden Markov model based speaker verification system.

• Phonetics and Speech Sciences
• /
• v.10 no.2
• /
• pp.7-14
• /
• 2018

The paper investigates the acoustic characteristics of English fricatives in the TIMIT corpus, with a special focus on the role of voicing in rendering fricatives in American English. The TIMIT database includes 630 talkers and 2,342 different sentences, and comprises more than five hours of speech. Acoustic analyses are conducted in the domain of spectral and temporal properties by treating gender, voicing, and place of articulation as independent factors. The results of the acoustic analyses revealed that acoustic signals interact in a complex way to signal the gender, place, and voicing of fricatives. Classification experiments using a multiclass support vector machine (SVM) revealed that 78.7% of fricatives are correctly classified. The majority of errors stem from the misclassification of /θ/ as [f] and /ʒ/ as [z]. The average accuracy of gender classification is 78.7%. Most errors result from the classification of female speakers as male speakers. The paper contributes to the understanding of the effects of voicing and gender on fricatives in a large-scale speech corpus.

• Speech Sciences
• /
• v.10 no.1
• /
• pp.71-84
• /
• 2003

In the acoustic modeling for large vocabulary speech recognition, a sparse data problem caused by a huge number of context-dependent (CD) models usually leads the estimated models to being unreliable. In this paper, we develop a new clustering method based on the C45 decision-tree learning algorithm that effectively encapsulates the CD modeling. The proposed scheme essentially constructs a supervised decision rule and applies over the pre-clustered triphones using the C45 algorithm, which is known to effectively search through the attributes of the training instances and extract the attribute that best separates the given examples. In particular, the data driven method is used as a clustering algorithm while its result is used as the learning target of the C45 algorithm. This scheme has been shown to be effective particularly over the database of low unknown-context ratio in terms of recognition performance. For speaker-independent, task-independent continuous speech recognition task, the proposed method reduced the percent accuracy WER by 3.93% compared to the existing rule-based methods.

• Speech Sciences
• /
• v.14 no.2
• /
• pp.105-114
• /
• 2007

The performance of speaker identifier is severely degraded in noisy environments. A study suggested the concept of observation membership for enhancing performances of speaker identifier with noisy speech [1]. The method scaled observation probabilities of input speech by observation identification values decided by SNR. In the paper [1], the authors suggested heuristic parameter values for membership function. In this paper we attempt to apply particle swarm optimization (PSO) for obtaining the optimal parameters for speaker identification in noisy environments. With the speaker identification experiments using the ETRI database we prove that the optimization approach can yield better performance than using only the original membership function.

• Phonetics and Speech Sciences
• /
• v.9 no.2
• /
• pp.85-94
• /
• 2017

Conventional speaker verification systems using time delay neural network, identity vector and probabilistic linear discriminant analysis (TDNN-Ivector-PLDA) are known to be very effective for verifying long-duration speech utterances. However, when test utterances are of short duration, duration mismatch between enrollment and test utterances significantly degrades the performance of TDNN-Ivector-PLDA systems. To compensate for the I-vector mismatch between long and short utterances, this paper proposes to use probabilistic linear discriminant analysis (PLDA) model adaptation with augmented data. A PLDA model is trained on vast amount of speech data, most of which have long duration. Then, the PLDA model is adapted with the I-vectors obtained from short-utterance data which are augmented by using vocal tract length perturbation (VTLP). In computer experiments using the NIST SRE 2008 database, the proposed method is shown to achieve significantly better performance than the conventional TDNN-Ivector-PLDA systems when there exists duration mismatch between enrollment and test utterances.