This paper investigated AV benefit of Korean and Japanese EFL learners' perception of the liquid contrast in English. In a perception experiment, the two English consonants /l/ and /r/ were embedded in initial and medial position in nonsense words in the context of the vowels /i, a, u/. Singletons and clusters were included in the speech material. Audio and video recordings were made using a total of 108 items. The items were presented to Korean and Japanese learners of English in three conditions: audio-alone (A), visual-alone (V) and audio-visual presentation (AV). The results showed that there was no evidence of AV benefit for the perception of the /l/-/r/ contrast for either Korean or Japanese learners of English. The results suggest that increasing auditory proficiency in identifying a non-native contrast is linked with an increasing proficiency in using visual cues to the contrast.

This study focused on comparing the following acoustic changes according to age and sex in adulthood: Fo, Jitter, Shimmer, and NHR. One hundred twenty Korean adults were divided into three age groups (20's, 50's, and 70's) and two sex groups (male and female). The subjects of this study performed three tasks: (1) sustained three vowels; (2) read on paragraph of 'Taking a Walk' (3) explained a picture. The data was analyzed using the MDVP of Multi-Speech. In the parameter of Fo, sex and age were influential factors. In the parameters of Jitter, Shimmer and NHR, the effect of sex and age was different in all three parameters. When the groups organized by sex were analyzed by age, the 20's group showed a statistical difference in all four parameters (Fo, Jitter, Shimmer, and NKR), when compared to the other two age ranges of 50's and 70's. We need to consider our standard parameter for the normal voice in the Korean elderly because the 50's and 70's age normal groups in our study are out of the current range of normal in MDVP.

We investigated whether the first syllable plays an important role in lexical access in Korean visual word recognition. To do so, one lexical decision task (LDT) and two form primed LDT experiments examined the nature of the syllabic neighborhood effect. In Experiment 1, the syllabic neighborhood density and the syllabic neighborhood frequency was manipulated. The results showed that lexical decision latencies were only influenced by the syllabic neighborhood frequency. The purpose of experiment 2 was to confirm the results of experiment 1 with form-primed LDT task. The lexical decision latency was slower in form-related condition compared to form-unrelated condition. The effect of syllabic neighborhood density was significant only in form-related condition. This means that the first syllable plays an important role in the sub-lexical process. In Experiment 3, we conducted another form-primed LDT task manipulating the number of syllabic neighbors in words with higher frequency neighborhood. The interaction of syllabic neighborhood density and form relation was significant. This result confirmed that the words with higher frequency neighborhood are more inhibited by neighbors sharing the first syllable than words with no higher frequency neighborhood in the lexical level. These findings suggest that the first syllable is the unit of neighborhood and the unit of representation in sub-lexical representation is syllable in Korea.

This study examined the effect of English-language experience on Korean- and Japanese-English late learners' production of English unstressed vowels in terms of four acoustic phonetic features: F0, duration, intensity and vowel reduction. The learners manifested some improvement with experience. The native-like attainment of a phonetic feature, however, was related to the phonological status of that feature in the speakers' native language. The results suggest that the extent to which the non-native speakers' production of English unstressed vowels improved with English-language experience varied as a function of their native language background.

The purpose of this study is to find some acoustic parameters on frequency domain to distinguish the Korean nasals, $/m,\;n,\;{\eta}/$ from each other. The new parameters are devised on the basis of LTAS (Long Term Average Spectrum). The maximum peak amplitude and the relevant formant frequency are measured in low and high frequency range, respectively. The frequency of spectral valley and its energy level are also obtained in the specific frequency range of the spectrum. Spectral slope, total energy value in specific frequency range, statistical distribution of spectral energy like centroid, skewness, and kurtosis are suggested as new parameters as well. The parameters that show statistically significant differences across nasals are summerized as follows. 1) in syllable initial positions: the total energy value from 1,500 to 2,200 Hz(zeroENG); 2) in syllable final positions: the peak amplitude of the first formant(peak1_a), the formant frequency with maximum peak amplitude from 4,000 to 8,000 Hz(peak2_f), the maximum peak amplitude of the formant frequency from 4,000 to 8,000 Hz(peak2_a), and the total energy value from 1,500 to 2,200 Hz(zeroENG).

The aim of the present study was to investigate a simile comprehension of children with and without mental retardation. Thirteen children with mental retardation and ten normal children with matched receptive vocabulary ability participated in the study. They were between five and eight years old in the age of receptive vocabulary. Fourteen picture plates were used and each plate consisted of three pictures. The findings were as fellows. First, children with mental retardation made significantly more errors than normal children in the comprehension of simile. Second, mental retarded children and normal children did not have a significant correlation between receptive vocabulary development and comprehension of simile. Finally, on simile interpretation type, children with mental retardation were likely to think a tenor into a vehicle because they failed to recognize the syntactic structure of simile, 'A is like B'. Thus, It is important to teach mental retarded children the syntactic structures of simile and to provide them with a variety of experience as well as to let them know an exact meaning of words.

The purpose of this study is to observe how Korean listeners detect a target phoneme with 'Focus' represented by prosodic prominence and question-induced semantic emphasis, and with intonational manipulation. According to the automated phoneme detection task using E-Prime, the Korean listeners detected phoneme targets more rapidly when the target-bearing words were in prominence position and in question-induced position. However, the presence of question-induced semantic emphasis reduced the prominence effect, so two effects interacted: when question-induced emphasis were primarily given as a cue, prominence which was given as secondary cue affected less to fine the new information. Besides, the intonation with manipulation was responded to faster than without manipulation.

Since computing the log-likelihood of Gaussian mixture density is a major computational burden for the speech recognizer based on the continuous HMM, several techniques have been proposed to reduce the number of mixtures to be used for recognition. In this paper, we propose a modified Bucket Box Intersection (BBI) algorithm, in which two relative thresholds are employed: one is the relative threshold in the conventional BBI algorithm and the other is used to reduce the number of the Gaussian boxes which are intersected by the hyperplanes at the boxes' edges. The experimental results show that the proposed algorithm reduces the number of Gaussian mixtures by 12.92% during the recognition phase with negligible performance degradation compared to the conventional BBI algorithm.

W3C announced a standard software architecture for multimodal context-aware middleware that emphasizes modularity and separates structure, contents, and presentation. We implemented a distributed multimodal interface system followed the W3C architecture, based on SCXML. SCXML uses parallel states to invoke both XHTML and VoiceXML contents as well as to gather composite or sequential multimodal inputs through man-machine interactions. We also hire Delivery Context Interface(DCI) module and an external service bundle enabling middleware to support context-awareness services for real world environments. The provision of personalized user interfaces for mobile devices is expected to be used for different devices with a wide variety of capabilities and interaction modalities. We demonstrated the implemented middleware could maintain multimodal scenarios in a clear, concise and consistent manner by some experiments.

This paper proposes class histogram equalization (CHEQ) to compensate noisy acoustic features for robust speech recognition. CHEQ aims to compensate for the acoustic mismatch between training and test speech recognition environments as well as to reduce the limitations of the conventional histogram equalization (HEQ). In contrast to HEQ, CHEQ adopts multiple class-specific distribution functions for training and test environments and equalizes the features by using their class-specific training and test distributions. According to the class-information extraction methods, CHEQ is further classified into two forms such as hard-CHEQ based on vector quantization and soft-CHEQ using the Gaussian mixture model. Experiments on the Aurora 2 database confirmed the effectiveness of CHEQ by producing a relative word error reduction of 61.17% over the baseline met-cepstral features and that of 19.62% over the conventional HEQ.

In this paper, we present several confidence measures (CM) for speech recognition systems to evaluate the reliability of recognition results. We propose heuristic CMs such as mean log-likelihood score, N-best word log-likelihood ratio, likelihood sequence fluctuation and likelihood ratio testing(LRT)-based CMs using several types of anti-models. Furthermore, we propose new algorithms to add weighting terms on phone-level log-likelihood ratio to merge word-level log-likelihood ratios. These weighting terms are computed from the distance between acoustic models and knowledge-based phoneme classifications. LRT-based CMs show better performance than heuristic CMs excessively, and LRT-based CMs using phonetic information show that the relative reduction in equal error rate ranges between $8{\sim}13%$ compared to the baseline LRT-based CMs. We use the support vector machine to fuse several CMs and improve the performance of utterance verification. From our experiments, we know that selection of CMs with low correlation is more effective than CMs with high correlation.

This paper describes a speech enhancement scheme that leads to significant improvements in recognition performance when used in the ASR front-end. The proposed approach is based on adaptive comb filtering and an MMSE-related parameter estimator. While adaptive comb filtering reduces noise components remarkably, it is rarely effective in reducing non-stationary noises. Furthermore, due to the uniformly distributed frequency response of the comb-filter, it can cause serious distortion to clean speech signals. This paper proposes an improved comb-filter that adjusts its spectral magnitude to the original speech, based on the speech absence probability and the gain modification function. In addition, we introduce the modified comb filtering-based speech enhancement scheme for ASR in mobile environments. Evaluation experiments carried out using the Aurora 2 database demonstrate that the proposed method outperforms conventional adaptive comb filtering techniques in both clean and noisy environments.

Discrimination between speech and music is important in many multimedia applications. Previously we proposed a new parameter for speech/music discrimination, the mean of minimum cepstral distances (MMCD), and it outperformed the conventional parameters. One weakness of MMCD is that its performance depends on range of candidate frames to compute the minimum cepstral distance, which requires the optimal selection of the range experimentally. In this paper, to alleviate the problem, we propose a multi-dimensional MMCD parameter which consists of multiple MMCDS with combination of different candidate frame ranges. Experimental results show that the multi-dimensional MMCD parameter yields an error rate reduction of 22.5% compared with the optimally chosen one-dimensional MMCD parameter.