• Phonetics and Speech Sciences
• /
• v.15 no.3
• /
• pp.9-16
• /
• 2023

The current study investigated the influence of age of L2 acquisition (AOA) and length of residence (LOR) in the L2 setting country on the production of voicing-conditioned vowel duration and spectral qualities in English by Korean learners. The primary aim was to explore the ways in which the language-specific phonetic features are acquired by the age of onset and L2 experience. Analyses of the archived corpus data produced by 45 native speakers of Korean showed that, regardless of AOA or LOR, absolute vowel duration was used as a salient correlate of voicing contrast in English for Korean learners. The accuracy of relative vowel duration was influenced more by onset age than by L2 experience, suggesting that being exposed to English at an early age may benefit the acquisition of temporal dimension. On the other hand, the spectral characteristics of English vowels were more consistently influenced by L2 experience, indicating that immersive experience in the L2 speaking environment are likely to improve the accurate production of vowel quality. The distinct influence of the onset age and L2 experience on the specific phonetic cues in L2 vowel production provides insight into the intricate relationship between the two factors on the manifestation of L2 phonological knowledge.

• Journal of the Institute of Convergence Signal Processing
• /
• v.4 no.4
• /
• pp.30-39
• /
• 2003

In this paper, we carried out the performance evaluation of HM-Net(Hidden Markov Network) speech recognition system for Korean speech databases. We adopted to construct acoustic models using the HM-Nets modified by HMMs(Hidden Markov Models), which are widely used as the statistical modeling methods. HM-Nets are carried out the state splitting for contextual and temporal domain by PDT-SSS(Phonetic Decision Tree-based Successive State Splitting) algorithm, which is modified the original SSS algorithm. Especially it adopted the phonetic decision tree to effectively express the context information not appear in training speech data on contextual domain state splitting. In case of temporal domain state splitting, to effectively represent information of each phoneme maintenance in the state splitting is carried out, and then the optimal model network of triphone types are constructed by in the parameter. Speech recognition was performed using the one-pass Viterbi beam search algorithm with phone-pair/word-pair grammar for phoneme/word recognition, respectively and using the multi-pass search algorithm with n-gram language models for sentence recognition. The tree-structured lexicon was used in order to decrease the number of nodes by sharing the same prefixes among words. In this paper, the performance evaluation of HM-Net speech recognition system is carried out for various recognition conditions. Through the experiments, we verified that it has very superior recognition performance compared with the previous introduced recognition system.

• MALSORI
• /
• no.57
• /
• pp.1-14
• /
• 2006

This study examines personal factors that affect Korean speakers' English pronunciation. Personal factors which are examined here are as follows: personality type, cognitive system, motivational orientation type, interest in English, how often they listen to tapes, and academic achievements. Data were collected through MBTI (Myers Briggs Type Indicator) Test, Group Embedded Figural Test, and a Questionnaire. The participants consisted of 65 college students. All the results were statistically analyzed: Korean students' personality type and cognitive system are not related with their pronunciation, but motivational orientation type, how often they listen to tapes, academic achievements, and interest in English study are correlated with their pronunciation.

• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.55-58
• /
• 2007

This study was investigated the effect of speech tasks on habitual pitch. Seven male and female young adult speakers participated in this study. The experiment consisted of seven different speech tasks: counting, reading, sustained phonation /a/, prolonged /i:/, answering /ne/. Data was analyzed via Visi-pitch IV. The results showed that there was no significant F0 difference among speech tasks.

• Proceedings of the KSPS conference
• /
• 2006.05a
• /
• pp.123-125
• /
• 2006

One of the biggest problems unsolved in emotional speech acquisition is how to make or find a situation which is close to natual or desired state from humans. We proposed a method to collect emotional speech data by scriptual context. Several contexts from the scripts of drama were chosen by the experts in the area. Context were divided into 6 classes according to the contents. Two actors, one male and one female, read the text after recognizing the emotional situations in the script.

• Speech Sciences
• /
• v.3
• /
• pp.93-108
• /
• 1998

This paper attempts to acoustically examine the place of articulation of Korean affricates. In order to pursue an acoustic analysis of where Korean affricates are articulated, we resort to LPC spectra of the Korean plain affricate /c/ in intervocalic position, based on theoretical assumptions (e.g., Stevens 1993a), and compare the data to that of the Korean alveolar consonants /t, s/ in the same context. Our phonetic results show that in intervocalic position, the Korean plain affricate is alveolar just like the Korean alveolar consonants /t, s/, supporting the articulatory studies of $Skali{\check{c}}kov{\acute{a}}$ (1960) and Kim (1997).

• Proceedings of the KSPS conference
• /
• 2000.07a
• /
• pp.211-219
• /
• 2000

In this paper, we examine fO peak lagging phenomenon in English. FO peak lagging refers to the fact that fO peak corresponding to an accent is realized beyond the domain of the host syllable. We present experimental data of fO peak lagging, which shows that fO peak is heavily delayed when the duration of the accented syllable is relatively short. In addition, we show that fO peak is also heavily delayed and realized in the following syllable in a focused word, even where the target vowel is not intrinsically short.

• Proceedings of the KSPS conference
• /
• 2003.05a
• /
• pp.131-134
• /
• 2003

This paper describes an efficient algorithm to generate compact and complete prompts lists for connected spoken digits database. In building a connected spoken digit recognizer, we have to acquire speech data in various contexts. However, in many speech databases the lists are made by using random generators. We provide an efficient algorithm that can generate compact and complete lists of digits in various contexts. This paper includes the proof of optimality and completeness of the algorithm.

• MALSORI
• /
• no.56
• /
• pp.207-223
• /
• 2005

The goal of language model adaptation is to improve the background language model with a relatively small adaptation corpus. This study presents a language model adaptation technique where additional text data for the adaptation do not exist. We propose the information retrieval (IR) technique with N-gram language modeling to collect the adaptation corpus from baseline text data. We also propose to use a dynamic language model interpolation coefficient to combine the background language model and the adapted language model. The interpolation coefficient is estimated from the word hypotheses obtained by segmenting the input speech data reserved for held-out validation data. This allows the final adapted model to improve the performance of the background model consistently The proposed approach reduces the word error rate by $13.6\%$ relative to baseline 4-gram for two-hour broadcast news speech recognition.

• Proceedings of the KSPS conference
• /
• 2003.10a
• /
• pp.149-152
• /
• 2003

In this paper, data-driven temporal filter methods[1] are investigated for robust feature extraction. A principal component analysis technique is applied to the time trajectories of feature sequences of training speech data to get appropriate temporal filters. We did recognition experiments on the Korean connected digit telephone speech database released by SITEC, with data-driven temporal filters. Experimental results are discussed with our findings.