Because the accuracy and the fluency are the essence in English speaking, both of them are very important in English trencher training and in-service English training programs. To get the accuracy and the fluency, the causes and the phenomena of the unnatural pronunciation have to be diagnosed. Consequently, the problematic and unnatural pronunciation of Korean elementary and secondary English teachers should be analyzed with using Acoustic Analyzing tools like CSL, Multi-speech and Praat. In addition, an attempt to Pinpoint what the causes of unnatural pronunciation was executed. Next a procedure and steps were proposed for in-service training programs that would cultivate the fluency and the accuracy. In case of elementary teachers, the unnatural pronunciation of segmental features and suprasegmental features were found much. therefore segmental features should be emphasized in the begging of pronunciation training courses and then suprasegmental features have to be emphasized. In case of secondary teachers, the unnatural pronunciation of suprasegmental features were found much. Therefore segmental and suprasegmental features have to be focused at the same time. In other words, features in word level should be focused first for elementary English teacher, and features in word level and beyond word level should be trained at the same time for secondary English teachers.

The purpose of this study is to investigate the production of English vowels by native speakers of Korean. In the production test, two English speakers and four native Korean speakers served as subjects. The four native Korean speakers were divided into two groups, experienced and inexperienced. Native English speakers generally showed significant differences both in vowel duration and in F1 & F2 values between members of vowel pairs which are of special interest of this study: /i/l vs. /I/, /$\varepsilon$/ vs. /${\ae}$/, and /u/ vs. /$\mho$/. The overall results showed that the experienced group produced more accurate results in vowel duration, F1, and F2 values.

This study was conducted to examine the effect of word length on auditory word recognition. Word length can be defined by several sublexical units, such as letters, phonemes, syllables, etc. To find out which sublexical units are influential in auditory word recognition, the auditory lexical decision task was used. In Experiment 1, we examined the partial correlation between the speed of reaction time and the number of sublexical units, and in Experiment 2, we executed ANOVA to find out which sublexical length variable was an influential unit. Through these two experiment, we concluded syllable length was the most important variable on auditory word recognition.

Mean F0, Jitter, Shimmer, and NHR were measured in 21 opera singers, 13 Korean traditional 'pansori' singers, and 21 non-singers. Four voice sampling conditions were ordinary /α/ prolongation, /α/ prolongation using vocal technique as if the subjects were singing, oral reading of the words of a song, and singing. One-way ANOVA and Sheffe post-hoc analysis were used to identify significant differences among the groups. The mean F0 was significantly different among the 3 groups in all conditions except for the ordinary /α/ prolongation condition. The opera singers produced lower jitter and NHR compared to 'pansori' singers and non-singers only in the singing condition.

The purpose of this research is to investigate how Japanese speakers perceive and produce Lenis, Aspirated and Fortis consonants in Korean. Identification tasks and production tasks were performed. The error analysis of both task showed that the participants had a significant difficulty in discriminating between Lenis and Aspirated sounds. And it was observed that there was a positive correlation between identification scores and production scores.

Channel distortion degrades the performance of speech recognizer in telephone environment. It mainly results from the bandwidth limitation and variation of transmission channel. Variation of channel characteristics is usually represented as baseline shift in the cepstrum domain. Thus undesirable effect of the channel variation can be removed by subtracting the mean from the cepstrum. In this paper, to improve the recognition performance of Korea connected digit telephone speech, channel compensation methods such as CMN (Cepstral Mean Normalization), RTCN (Real Time Cepatral Normalization), MCMN (Modified CMN) and MRTCN (Modified RTCN) are applied to the static MFCC. Both MCMN and MRTCN are obtained from the CMN and RTCN, respectively, using variance normalization in the cepstrum domain. Using HTK v3.1 system, recognition experiments are performed for Korean connected digit telephone speech database released by SITEC (Speech Information Technology & Industry Promotion Center). Experiments have shown that MRTCN gives the best result with recognition rate of 90.11% for connected digit. This corresponds to the performance improvement over MFCC alone by 1.72%, i.e, error reduction rate of 14.82%.

Speech recognition performance is severly degraded under noisy envrionments. One approach to cope with this problem is audio-visual speech recognition. In this paper, we discuss the experiment results of bimodal speech recongition based on enhanced speech feature vectors using lip information. We try various kinds of speech features as like linear predicion coefficient, cepstrum, log area ratio and etc for transforming lip information into speech parameters. The experimental results show that the cepstrum parameter is the best feature in the point of reconition rate. Also, we present the desirable weighting values of audio and visual informations depending on signal-to-noiso ratio.

There are many speech analysis tools available. Among them real-time analysis tool is very useful for interactive experiments. A real-time speech analysis tool was implemented using Matlab. Matlab is a very widely used general purpose signal processing tool. In general, its computational speed is relatively lower than that of the codes from conventional programming languages. Especially, real-time analysis including input of signal and output of the result was not possible in the past. However, due to the improvement of computing power of PCs and inclusion of real-time I/O toolboxes in Matlab, real-time analysis is now possible in some extent by Matlab only. In this experiment, we tried to implement a real-time speech analysis tool using Matlab. Pitch and spectral information is computed in real-time. From the result it is shown that such real-time applications can be implemented easily using Matlab.

In Korean, each digit is monosyllable and some pairs are known to have high confusability, causing performance degradation of connected digit recognition systems. To improve the performance, in this paper, we employ various discriminant analyses (DA) including Linear DA (LDA), Weighted Pairwise Scatter LDA WPS-LDA), Heteroscedastic Discriminant Analysis (HDA), and Maximum Likelihood Linear Transformation (MLLT). We also examine several combinations of various DA for additional performance improvement. Experimental results show that applying any DA mentioned above improves the string accuracy, but the amount of improvement of each DA method varies according to the model complexity or number of mixtures per state. Especially, more than 20% of string error reduction is achieved by applying MLLT after WPS-LDA, compared with the baseline system, when class level of DA is defined as a tied state and 1 mixture per state is used.

Voice input is often required in many new application environments, but the low rate of speech recognition makes it difficult to extend its application. Previous approaches were to raise the accuracy of the recognition by post-processing of the recognition results, which were all lexical-oriented. We suggest a new semantic-oriented approach in speech recognition error correction. Through experiments using a speech-driven in-vehicle telematics information application, we show the excellent performance of our approach and some advantages it has as a semantic-oriented approach over a pure lexical-oriented approach.

Music query by humming is a challenging problem since the humming signal inevitably contains much variation and inaccuracy. In this paper, we suggest an algorithm for querying a wanted song from music database by humming its melody. In order to suit or adapt the inaccurate peoples humming, a new melody representation technique is proposed. Our algorithm is basically a pitch and duration information-based one and performs fairly well. 85% of correct query rate of the song is achieved for the top 3 matches when tested with 20 songs.

This paper describes the reduction of DB without degradation of speech quality in Corpus-based Speech synthesizer of the Korean language. In this paper, it is proposed that the frequency of every unit in reduced DB reflect the frequency of units in the Korean language. So, the target population of every unit is set to be proportional to its frequency in Korean large corpus (780k sentences, 45Mega phones). Secondly, the frequent instances during synthesis should be also maintained in reduced DB. To the last, it is proposed that frequency of every instance be reflected in clustering criteria and used as another important criterion for selection of representative instances. The evaluation result with proposed methods reveals better quality than that using conventional methods.

We propose a weighted filter bank analysis and model adaptation (WFBA-MA) scheme to improve the utilization of uncorrupted or less severely corrupted frequency regions for robust speech recognition. A weighted met frequency cepstral coefficient is obtained by weighting log filter bank energies with reliability coefficients and hidden Markov models are also modified to reflect the local reliabilities. Experimental results on TIDIGITS database corrupted by band-limited noises and car noise indicated that the proposed WFBA-MA scheme utilizes the uncorrupted speech information well, significantly improving recognition performance in comparison to multi-band speech recognition systems.