This study investigated the reading ability of Korean second-graders and the relationship between reading and phonological awareness and RAN (Rapid Automatized Naming) ability. A language-based reading assessment battery was used. Children at the end of the Korean second-grade were still at the developmental stage of decoding skill and seemed to be at Chall's stage 1. Findings indicated significant correlations between reading ability and phonological awareness and between reading ability and RAN ability. Therefore, the importance of phonological processing could be extended to syllable-based alphabetic languages.

The purpose of the present study was to investigate the characteristics of metalinguistic awareness of children with specific language impairment. Forty-five children participated in this study; 15 children with specific language impairment (SLI group) whose range of the language age was 4;6-6;6, 15 normal children chronological age matched (CA controls) and 15 normal children language age matched with the SLI group (CA controls). A metalinguistic task involving the identification and revision of syntactic, semantic, and phonologic errors was used. The SLI group performed significantly poorer than CA controls as well as LA controls in identifying and correcting error sentences, especially sentences with syntactic error. These results revealed the relation between language problems of children with SLI and metalinguistic abilities.

Differential instrumentation for the diagnoses of individuals with Cleft palate has been used to objectively measure speech problems. The Cepstrum Method was used to study the vocal tract transfer function. The vocal tract transfer function and the source spectrum should be considered in the evaluation of nasal resonance. The aim of this study was to collect quantitative data on the acoustic Instrumentation used for evaluating hypernasality. Normal subjects (9 male, 21 female; 37 male children, 20 female children) and individuals with VPI (13 male, 8 female; 16 male children, 9 female) participated in this study. The vowel /i/ was selected to gauge the severances of hypernasality Spectral and Cepstral studies using CSL was used to identify the acoustic characteristics. Cepstrum analysis shows significant differences in quefrency and amplitude. The quefrency of normal groups was shorter than that of the VPI groups, while the amplitude of normal groups was lower than that of the VPI groups. This may have significance in the evaluation 'of nasal resonance.

The phenomenon of vocal vibrato may be regarded as an acoustic representation of one of the most rapid and continuous changes in pitch and intensity that the human vocal mechanism is capable of producing. Singers are likely to use vibrato effectively to enrich their voice. The purpose of this study was to obtain acoustic measurements (vF0 and vAm) of 45 subjects (15 trot and 15 ballad singers and 15 non-singers) and to compare acoustic measurements of the vowel /a/ produced by 3 groups on 2 voice sampling conditions (prolongation and singing of /a/). Thirty singers of trot and ballad were selected by a producer and a concert director working for the KBS (Korean Broadcasting System). The MDVP was used to measure the acoustic parameters. A two-way MANOVA was used for statistical analyses. The results were as follows; Firstly, there was no significant difference among the 3 groups in vF0 and vAm in prolongation of /a/, but in singing voice, there was a significant difference among 3 groups in vF0 and vAm. Secondly, there was an interaction between music genre and voice sampling condition in vF0, and vAm. Finally, trot singers sing with more vibrato than ballad singers. It was concluded that it is very important to analyze singers' voice including various voice conditions (prolongation, reading, conversation, and singing) and to identify differences of singing voice characteristics among music genre.

The purpose of this study is to compare disfluencies and speech rates (overall speech rate and articulation rate) of normal adult speakers who use the standard Korean according to dissimilar speech tasks (story-telling and text-reading). Participants were 100 Korean adult speakers. The results are summarized as follows: First, the most frequent type of disfluency in the story-telling task was 'interjection', whereas that in the text-reading task was 'revision'. Second, the overall speech rates (syllables per second and syllables per minute) showed significant differences depending on the speech tasks. Third, the articulation rates (syllables per second and syllables per minute) showed significant differences depending on the speech tasks.

This paper introduces a 'Hanmal' Korean language diphone database for speech synthesis, which has been publicly available since 1999 in the MBROLA web site and never been properly published in a journal. The diphone database is compatible with the MBROLA programme of high-quality multilingual speech synthesis systems. The usefulness of the diphone database is introduced in the paper. The paper also describes the phonetic and phonological structure of the database, showing the process of creating a text corpus. A machine-readable Korean SAMPA convention for the control data input to the MBROLA application is also suggested. Diphone concatenation and prosody manipulation are performed using the MBR-PSOLA algorithm. A set of segment duration models can be applied to the diphone synthesis of Korean.

The aim of this paper is to suggest an index of coarticulation, proportional undershoot values, given the observation that absolute undershoot within a language tends to be proportional to target-locus difference. The target-locus proportionality predicts that a large difference between the consonant locus and the vowel target will result in a large amount of vowel undershoot, while a small difference a small amount of vowel undershoot. It turns out that the proportional undershoot is a potentially more appropriate way of comparing degree of undershoot across languages. However, even though the proportional undershoot measurement may provide a useful index comparing the overall coarticulation degree in a CV token for cross-linguistic data, it is concluded that it may potentially wrongly predict the cases of transfer or error as a progress in learning.

In this study I investigate how native Korean listeners perceive English vowels /i/, /I/, and /$\epsilon$/. I extend Flege et al's (1997) study with synthesized /i/-/I/ and /I/-/$\epsilon$/ continua, and apply the results to Flege's (1995) Speech Learning Model (SLM). The statistical results show that native speakers of English rely more on spectral steps than on vowel duration when they identify the /i/-/I/ continuum, whereas native speakers of Korean rely more on vowel duration than on spectral steps when they identify the same continuum. In the case of the /I/-/$\epsilon$/ continuum, both groups rely on spectral steps when they identify the /$\epsilon$/, which supports the SLM; Koreans identified the /$\epsilon$/ categorically since Korean has the equivalent vowel. However, there was not statistical difference between Korean subjects with more English experience (KE) and those with less English experience in the identification of both continua. This contradicts the SLM, which posits that experienced L2 learners are better than inexperienced L2 learners in perception of L2 sounds. The exact nature of this should be further investigated in the SLM.

This paper is a production study of the effects of Korean prosody on two voiceless coronal fricatives /$s^h$/ and /s*/. The target segments were embedded in four prosodic positions: initial to the Intonational Phrase or the Accentual Phrase, and medial to the Accentual Phrase or to the Prosodic Word. Acoustic measurements showed that the durational differences associated with the /$s^h$/ versus /s*/ contrast vary in magnitude in different prosodic positions, confirming the proposal that segmental properties are affected by prosodic categories. This suggests that any speech synthesizer should take into consideration prosodically conditioned durational variation.

The current study considers that the distinctive acoustic properties and variations in the closed plateau are realized by four different pitch accents(/Hi*_ H*/ or /Hi*_$h*_f$ for an emphatic phrase and /hi*_ H*/ for a neutral phrase) in an intermediate phrase in the French intonational structure. Thus, an attempt is made to define the acoustic property of the CP in the ip according to the duration time and pitch range, while different combinations of the four pitch accents of the CP are used to explain the way a speaker will highlight. The duration time of the CP was measured at about 0.67 sec. for males and 0.75 sec. for females. The duration properties of the plateau in the CP were found to control the pitch range based on two different prominent pitch accents, which appeared in more than two APs. Therefore, the ip was identified as having a hierarchical level in the French intonational structure, along with the AP and IP. In addition, the CP in the ip was used as a specific location to explain the pragmatic meaning of the rhythm using the two acoustic factors and different combinations of the four pitch accents.

Confidence measure (CM) is used for the rejection of mis-recognized words in an automatic speech recognition (ASR) system. Rahim, Lee, Juang and Cho's confidence measure (RLJC-CM) is one of the widely-used CMs [1]. The RLJC-CM is calculated by averaging phone-level CMs. An extension of the RLJC-CM was achieved by Kim et al [2]. They devised the normalized CM (NCM), which is a statistically normalized version of the RLJC-CM by using the tri-phone based CM normalization. In this paper we verify the NCM by generalizing tri-phone to n-phone unit. To apply various units for the normalization, mono-phone, tri-phone, quin-phone and $\infty$-phone are tested. By the experiments in the domain of the isolated word recognition we show that tri-phone based normalization is sufficient enough to enhance the rejection performance of the ASR system. Also we explain the NCM in regard to two class pattern classification problems.

An important aspect of video indexing is the ability to segment video into meaningful segments, i.e., content-based video segmentation. Since the audio signal in the sound track is synchronized with image sequences in the video program, a speech signal in the sound track can be used to segment video into meaningful segments. In this paper, we propose a new approach to content-based video segmentation. This approach uses closed caption to construct a recognition network for speech recognition. Accurate time information for video segmentation is then obtained from the speech recognition process. For the video segmentation experiment for TV news programs, we made 56 video summaries successfully from 57 TV news stories. It demonstrates that the proposed scheme is very promising for content-based video segmentation.