This paper criticizes S. Kim (1992), claiming that the perception of place cues in the reduced stop consonant clusters ('reducing' means 'cutting off' the acoustic silence in stop clusters) largely depends on the acoustic characteristics such as formant transition and noise frequency distribution of stop burst, rather than the closure duration time as advocated by S. Kim (1992). The claim is based on the perception test conducted upon 111 stimuli over 10 subjects. The finding is that, when the closure duration is cut off up to the point where only one stop is perceived, place of the second stop, not the first one, in the cluster is in most cases perceived regardless of the places of the first and second stops. It is likely that the place cues of the stop in the prevocalic position mask those in the postvocalic position.

This study is a kind of a preliminary study to develop naturalness of dialog TTS system. In this study, as major characteristics of politeness in Korean, temporal(total duration of utterances, speech rate and duration of utterance final syllables) and F0(mean F0, boundary tone pattern, F0 range) features were discussed through acoustic analysis of recorded data of semantically neutral sentences, which were spoken by ten professional voice actors under two conditions of utterance type - namely, normal and polite type. The results show that temporal characteristics were significantly different according to the utterance type but F0 characteristics were not.

This is a study on perception errors of Korean stops in word initial position. This study will show how listeners perceive manipulated Korean stops through two experiments and I will analyze why they are confused. In addition, 1 will show those perception errors not only reflect the relationship between the place and manner of articulation but also ate deeply related to the length of aspiration.

Laryngeal diseases cause significant changes in the quality of speech production. Automatic detection of laryngeal diseases by voice is attractive because of its nonintrusive nature. In this paper, we apply speech recognition techniques to detection of laryngeal cancer, and investigate which feature parameters and classification methods are appropriate for this purpose. Linear Predictive Cepstral Coefficients (LPCC) and Mel-Frequency Cepstral Coefficients (MFCC) are examined as feature parameters, and parameters reflecting the periodicity of speech and its perturbation are also considered. As for classifier, multilayer perceptron neural networks and Gaussian Mixture Models (GMM) are employed. According to our experiments, higher order LPCC with the periodic information parameters yields the best performance.

In this paper, we consider the implementation of connected digit recognition system and the several approaches to improve its performance. To implement efficiently the fixed or variable length digit recognition system, finite state network (FSN) is required. We merge the word network algorithm that implements the FSN with one pass dynamic programming search algorithm that is used for general speech recognition system for fast search. To find the efficient modeling of digit recognition system, we perform some experiments along the various conditions to affect the performance and summarize the results.

The purpose of voice color conversion is to change the speaker identity perceived from the speech signal. In this paper, we propose a new voice color conversion algorithm through the formant shifting and the spectrum-tilt modification in the frequency domain. The basic idea of this technique is to convert the positions of source formants into those of target speaker's formants through interpolation and decimation and to modify the spectrum-tilt by utilizing the information of both speakers' spectrum envelops. The LPC spectrum is adopted to evaluate the position of formant and the information of spectrum-tilt. Our algorithm enables us to convert the speaker identity rather successfully while maintaining good speech quality, since it modifies speech waveforms directly in the frequency domain.

Utterance verification aims at rejecting both out-of-vocabulary (OOV) utterances and low-confidence-scored in-vocabulary (IV) utterances. For utterance verification on Korean connected digit recognition task, we investigate several methods to construct filler and anti-digit models. In particular, we propose a substitution error correction method based on 2-best decoding results. In this method, when 1st candidate is rejected, 2nd candidate is selected if it is accepted by a specific hypothesis test, instead of simply rejecting the 1st one. Experimental results show that the proposed method outperforms the conventional log likelihood ratio (LLR) test method.

In this paper, we present a single framework based on GMM for speaker recognition. The proposed framework can simultaneously minimize environmental variations on mismatched conditions and adapt the bias free and speaker-dependent characteristics of claimant utterances to the background GMM to create a speaker model. We compare the closed-set speaker identification for conventional method and the proposed method both on TIMIT and NTIMIT. In the several sets of experiments we show the improved recognition rates on a simulated channel and a telephone channel condition by 7.2% and 27.4% respectively.

An efficient method that compensates the HMM parameters for the noisy speech recognition is proposed. Instead of assuming some analytical approximations as in the PMC, the proposed method directly re-estimates the HMM parameters by the segmental k-means algorithm. The proposed method has shown improved results compared with the conventional PMC method at reduced computational cost.

Decision tree-based state tying has been proposed in recent years as the most popular approach for clustering the states of context-dependent hidden Markov model-based speech recognition. The aims of state tying is to reduce the number of free parameters and predict state probability distributions of unseen models. But, when doing state tying, the size of a decision tree is very important for word independent recognition. In this paper, we try to construct optimized decision tree based on the average of feature vectors in state pool and the number of seen modes. We observed that the proposed optimal decision tree is effective in predicting the state probability distribution of unseen models.