This study explored the perception of transplanted Korean prosody by NE (Native speakers of English) and NK (Native speakers of Korean) listeners. The Korean utterances of various sentence types produced by NE and NK were employed to transplant the original Korean prosody contours to the Korean utterances read by NE. Then, other NE and NK were instructed to rate the transplanted prosodic components. Results showed that the interactions between the two rater groups with the three factors (e.g., transplantation types & rater groups, sentence types & rater groups, sentence length & rater groups) turned out to be meaningful. Both rater groups preferred the combined effect of transplanted prosodic components (e.g. DP, DPI) to that of individual transplantation (e.g. I, D, P). Compared to NK, NE were more sensitive to duration change than pitch change whereas NK showed equal preference to the both. In sentence types such as De, Ex, Im, and Ta, NE perceived higher similarity than NK.

This paper attempts to analyze the characteristics of Korean vowel production by 12 Korean native speakers and 36 adult learners. The analyses have been performed with investigations of F1and F2 values. Results showed that there's no significant difference between /ㅔ/ and /H/ and between /ㅗ/ and /ㅜ/ in Korean native speakers' pronunciations. The distinguishing tendencies found in the analyses of foreign learners' pronunciations are fronting and lowering of /ㅗ/ by English speakers, backing and heightening of /ㅓ/ by Japanese speakers and backing and lowering of /ㅏ/ by Chinese speakers. For the limitations of this paper, it has a meaning of a preliminary study and could be developed into further research to show the order of acquisition and L1 transference.

The purposes of this experiment were to compare and analyze some voice parameters of the hearing impaired male adults and to suggest a basic data on the speech intervention for the hearing impaired. Voice analysis of four sustained vowels(/a/, /i/, /${\partial}$/, /u/, fundamental Sequency(F0), jitter percent, shimmer percent, and Noise to Harmonic Ratio(NHR) was conducted for the deaf young male adults using a sign laguage(N=5, aged 16-20) and the normal hearing young male adults(N=10, aged 18-20) by using MDVP(Multi-Dimensional Voice Program) in CSL. F0, jitter, and shimmer in the deaf group were significantly higher than those in the normal hearing group. The average of F0 was 151 Hz, which was lower than the results of the previous studies, and there were no significant differences among the sustained vowels. In both groups, the values of the voice parameters were stable on the /a/ or /${\partial}$/, those closed to the standard scores.

The aim of this study is to compare ASD (Autism Spectrum Disorder) children with normal children in terms of the prosodic features. Materials are collected by the reading of Korean read sentences. They are composed of 10 declarative sentences, each of which was consisted of 5-6 words. Subjects are consisted of 10 ASD and 10 normal male children with a receptive vocabulary age of 5;0-6;5 years. We found out that both groups showed the differences not only in the tonal patterns at the end of the prosodic phrases, but also in both the degree of rising and falling slope related to pitch contour. While HL% and HLH% were highly emerged in sentence final position in normal group, HL% and HLH% were prominent in ASD group in the same position. LH% and LHL% IP types were observed only in ASD group in sentence medial position. The slope showing the variation in the fundamental frequency at the end of the prosodic phrase was twice as steep in the group of ASD children as in the group of normal children.

We have been building a text-independent speaker recognition system that is robust to unknown channel and noise environments. In this paper, we propose a linear transformation to obtain robust features. The transformation is optimized to maximize the distances between the Gaussian mixtures. We use rotation of the axes, to cope with the problem of scaling the transformation matrix. The proposed transformation is similar to PCA or LDA, but can achieve better result in some special cases where PCA and LDA can not work properly. We use YOHO database to evaluate the proposed method and compare the result with PCA and LDA. The results show that the proposed method outperforms all the baseline, PCA and LDA.

In this paper, we propose a new variable bit-rate speech coder based on the waveform interpolation concept. After the coder extracted all parameters, the amounts of the distortions between the current and the predicted parameters which are estimated by extrapolation using past two parameters are measured for all parameters. A parameter would not be transmitted unless the distortion exceeds the preset threshold. At the decoder side, the non-transmitted parameter is reconstructed by extrapolation with past two parameters used to synthesize signals. In this way, we can reduce 26% of the total bit rate while retaining the speech quality degradation below 0.1 PESQ score.

In this paper, we propose a method for optimizing a multiple pronunciation dictionary used for modeling pronunciation variations of non-native speech. The proposed method removes some confusable pronunciation variants in the dictionary, resulting in a reduced dictionary size and less decoding time for automatic speech recognition (ASR). To this end, a confusability measure is first defined based on the Levenshtein distance between two different pronunciation variants. Then, the number of phonemes for each pronunciation variant is incorporated into the confusability measure to compensate for ASR errors due to words of a shorter length. We investigate the effect of the proposed method on ASR performance, where Korean is selected as the target language and Korean utterances spoken by Chinese native speakers are considered as non-native speech. It is shown from the experiments that an ASR system using the multiple pronunciation dictionary optimized by the proposed method can provide a relative average word error rate reduction of 6.25%, with 11.67% less ASR decoding time, as compared with that using a multiple pronunciation dictionary without the optimization.

This paper presents the effects of different phone-like-unit (PLU) sets in order to propose an optimal PLU set for the performance improvement of Korean automatic speech recognition (ASR) systems. The examination of 9 currently used PLU sets indicates that most of them include a selection of allophones without any sufficient phonetic base. In this paper, a total of 34 PLU sets are designed based on Korean phonetic characteristics arid the effects of each PLU set are evaluated through experiments. The results show that the accuracy rate of each phone is influenced by different phonetic constraint(s) which determine(s) the PLU sets, and that an optimal PLU set can be anticipated through the phonetic analysis of the given speech data.

In this work, we propose an automatic word spacing system for Korean using conditional random fields (CRFs) with Korean features. We map a word spacing problem into a classification problem in our work. We build a basic system which uses CRFs and Eumjeol bigram. After then, we analyze the result of inner-test. We extend a basic system added by some Korean features which are Josa, Eomi and two head Eumjeols of word extracting from lexicon. From the results of experiment, we can see that the proposed method is better than previous methods. Additionally the proposed method will be able to use mobile and speech applications because of very small size of model.

This paper focuses on the real time implementation of an adaptive multi-rate (AMR) speech codec, that is a standard speech codec of IMT-2000, using the TMS320VC5510. The series of TMS320VC55x is a 16-bit fixed-point digital signal processor (DSP) having low power consumption for the use of mobile communications by Texas Instruments (TI) corporation. After we analyze the AMR algorithm and source code as well as the structure and I/O of 7MS320VC55x, we carry out optimizing the programs for real time implementation. The implemented AMR speech codec uses 55.2 kbyte for the program memory and 98.3 kbyte for the data memory, and it requires 709,878 clocks, i.e. about 3.5 ms, for processing a frame of 20 ms speech signal.

We propose a new method to apply shape analysis techniques to a computational auditory scene analysis (CASA)-based speech separation system. The conventional CASA-based speech separation system extracts speech signals from a mixture of speech and noise signals. In the proposed method, we complement the missing speech signals by applying the shape analysis techniques such as labelling and distance function. In the speech separation experiment, the proposed method improves signal-to-noise ratio by 6.6 dB. When the proposed method is used as a front-end of speech recognizers, it improves recognition accuracy by 22% for the speech-shaped stationary noise condition and 7.2% for the two-talker noise condition at the target-to-masker ratio than or equal to -3 dB.