• Phonetics and Speech Sciences
• /
• v.1 no.1
• /
• pp.55-62
• /
• 2009

This paper proposes an improved method to verify a keyword segment that results from a word spotting system. First a baseline word spotting system is implemented. In order to improve performance of the word spotting systems, we use a two-pass structure which consists of a word spotting system and an utterance verification system. Using the basic likelihood ratio test (LRT) based utterance verification system to verify the keywords, there have been certain problems which lead to performance degradation. So, we propose a method which uses phone-level log-likelihood ratios (PLLR) patterns in computing confidence measures for each keyword. The proposed method generates weights according to the PLLR patterns and assigns different weights to each phone in the process of generating confidence measures for the keywords. This proposed method has shown to be more appropriate to word spotting systems and we can achieve improvement in final word spotting accuracy.

• ETRI Journal
• /
• v.27 no.4
• /
• pp.461-464
• /
• 2005

Recently, a variety of confidence measures for utterance verification has been studied to improve speech recognition performance by rejecting out-of-vocabulary inputs. Most of the conventional confidence measures for utterance verification are based primarily on hypothesis testing or an approximated posterior probability, and their performances depend on the robustness of an alternative hypothesis or the prior probability. We introduce a novel confidence measure called a search confusion rate (SCR), which does not require an alternative hypothesis or the approximation of posterior probability. Our confusion-based approach shows better performance in additive noise-corrupted speech as well as in clean speech.

• Speech Sciences
• /
• v.1
• /
• pp.137-152
• /
• 1997

The objectives of this paper are to see how the declination is realized in the different positions/lengths of the utterance, to see if the $F_0$ value throughout the utterance changes in a predictable way, and if so, to find out the best quantitative model which fits the declination. The experiment results are as follows. First, the peak value over the utterance can be affected by the position of the peak and length of the utterance. Second, the choice of quantitative models is dependent on the different list lengths. Third, in everyone's speech, there is a baseline (the lowest $F_0$ value a speaker can use), and the $F_0$ will not fall below the baseline. Forth, the peak $F_0$ of the last word in each list shows little variation in pitch value (target $F_0$) while the number of words in the list affects the starting $F_0$ values.

• Speech Sciences
• /
• v.13 no.3
• /
• pp.83-95
• /
• 2006

This study attempts to suggest a criterion of Korean language development. For this purpose we investigated speaking rates of the spontaneous utterances produced by 144 children, aged 3 to 8. We analyzed each subject's speaking rate and its relevance with speaker's age, gender and utterance length. To determine the relative contributions of variables to the speaking rate, multiple regression was conducted. Results of this study can be summarized as follows: (1) The mean and maximum values of the speaking rate increased with the growth of age. (2) A statistically significant increase in speaking rate appeared at two-year intervals. (3) There was no significant difference between male and female groups in the speaking rate. (4) The multiple regression analysis has shown that along with the speaker's age, the utterance length(the mean number of syllables per utterance) is also important in estimating the speaking rates.

• The Journal of the Acoustical Society of Korea
• /
• v.35 no.6
• /
• pp.501-509
• /
• 2016

We propose a method to improve the robustness of speaker verification on short test utterances. The accuracy of the state-of-the-art i-vector/probabilistic linear discriminant analysis systems can be degraded when testing utterance durations are short. The proposed method compensates for utterance variations of short test feature vectors using deep neural networks. We design three different types of DNN (Deep Neural Network) structures which are trained with different target output vectors. Each DNN is trained to minimize the discrepancy between the feed-forwarded output of a given short utterance feature and its original long utterance feature. We use short 2-10 s condition of the NIST (National Institute of Standards Technology, U.S.) 2008 SRE (Speaker Recognition Evaluation) corpus to evaluate the method. The experimental results show that the proposed method reduces the minimum detection cost relative to the baseline system.

• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.193-196
• /
• 2007

The purpose of this paper is to find boundary tone's characteristics in Korean emotion speeches. I mainly focus on investigating patterns and f0 values of boundary tones and f0 values in utterance final phrases.

• Proceedings of the KSPS conference
• /
• 1996.10a
• /
• pp.554-559
• /
• 1996

An utterance is normally divided into two or more intonation groups. Bach intonation group has its intonation pattern. Pitch movement of Spanish utterance is basically determined by a combination of two factors: position of the stressed syllables and the intonation pattern. The pitch of a syllable can be affected by that of preceding syllables. This is rather a physiological effect than a phonological one.

• Phonetics and Speech Sciences
• /
• v.12 no.2
• /
• pp.63-72
• /
• 2020

This study aims to develop a video-based smart utterance deep analyser (SUDA) application that analyzes semiautomatically the utterances that child and mother produce during interactions over time. SUDA runs on the platform of Android, iPhones, and tablet PCs, and allows video recording and uploading to server. In this device, user modes are divided into three modes: expert mode, general mode and manager mode. In the expert mode which is useful for speech and language evaluation, the subject's utterances are analyzed semi-automatically by measuring speech and language factors such as disfluency, morpheme, syllable, word, articulation rate and response time, etc. In the general mode, the outcome of utterance analysis is provided in a graph form, and the manger mode is accessed only to the administrator controlling the entire system, such as utterance analysis and video deletion. SUDA helps to reduce clinicians' and researchers' work burden by saving time for utterance analysis. It also helps parents to receive detailed information about speech and language development of their child easily. Further, this device will contribute to building a big longitudinal data enough to explore predictors of stuttering recovery and persistence.

• Phonetics and Speech Sciences
• /
• v.13 no.1
• /
• pp.53-64
• /
• 2021

According to Demand and Capacity Model (DCM), external and internal environments influence the disfluency of children who stutter (CWS). This study investigated the effects of simultaneous changes in motoric and linguistic demands on CWS and children who do not stutter (CWNS). Participants were 4-6 years old CWS and CWNS. A sentence imitation task with changes in speech rate, utterance length, and sentence complexity was used to examine their effects on children's disfluency. When the utterance length changed, CWS showed more disfluency regardless of utterance length and as the speech rate changed, CWS showed more disfluency at fast speech rate than CWNS. When the utterance length and speech rate changed, at fast speech rate, CWS showed more disfluency in both utterances than CWNS. When sentence complexity changed, CWS showed more disfluency than CWNS in complex sentences. Changes in linguistic elements such as speech rate, utterance length, and sentence complexity affect disfluency in CWS, especially when they were exposed to faster, longer, and more complex sentences. This indicates that CWS are vulnerable to fast and complex speech motor control and language processing ability than CWNS. Thus, this study suggests that parents and therapists consider both the speech rate and the utterance length when talking with CWS.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.11 no.8
• /
• pp.3021-3027
• /
• 2010

Example-based chatBot generates a response to user's utterance by searching the most similar utterance in a collection of dialogue examples. Though finding an appropriate example is very important as it is closely related to a response quality, few studies have reported regarding what features should be considered and how to use the features for similar utterance searching. In this paper, we propose a machine learning framework which uses various linguistic features. Experimental results show that simultaneously using both semantic features and lexical features significantly improves the performance, compared to conventional approaches, in terms of 1) the utilization of example database, 2) precision of example matching, and 3) the quality of responses.