• 말소리와 음성과학
• /
• 제2권3호
• /
• pp.11-22
• /
• 2010

The present study attempts to investigate declination in Korean and its relevant aspects to the length of utterance, the clause boundary, and focus. More specifically, I examine the relation of declination with the length of utterance, the declination reset at the clause boundary, and the effect of focus on declination. Results showed that the length of utterance had no relation with the first and last pitch values of the utterance but that they were consistent regardless of the length of utterance. However, the declination slope changed to be relatively gentle from the fourth accentual phrase to the end of the whole intonational phrase. There was a reset of declination in such a way that the first pitch in the second phrase was always lower than that of the first phrase, but the first pitch in the third phrase was not always lower than that of the second phrase when the whole utterance was composed of three phrases. Finally, the pitch values of the focusing words decreased as their position went back in a sentence. One declination line was formed in the case of focused utterance, but in the case of an utterance that contained a clause boundary, a new declination line was formed at the start of each new clause. These findings can be applied to developing a Korean speech synthesizer that contains natural prosody; they can be also utilized for teaching Korean prosody.

• 대한음성학회지:말소리
• /
• 제45호
• /
• pp.15-22
• /
• 2003

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.

• The Journal of the Acoustical Society of Korea
• /
• 제24권4E호
• /
• pp.144-151
• /
• 2005

We evaluate the performance of emotion recognition via speech signals when a plain speaker talks to an entertainment robot. For each frame of a speech utterance, we extract the frame-based features: pitch, energy, formant, band energies, mel frequency cepstral coefficients (MFCCs), and velocity/acceleration of pitch and MFCCs. For discriminative classifiers, a fixed-length utterance-based feature vector is computed from the statistics of the frame-based features. Using a speaker-independent database, we evaluate the performance of two promising classifiers: support vector machine (SVM) and hidden Markov model (HMM). For angry/bored/happy/neutral/sad emotion classification, the SVM and HMM classifiers yield $42.3\%\;and\;40.8\%$ accuracy, respectively. We show that the accuracy is significant compared to the performance by foreign human listeners.

• 대한음성학회지:말소리
• /
• 제56호
• /
• pp.173-183
• /
• 2005

We propose a speech emotion recognition method for affective human-robot interface. In the Proposed method, emotion is classified into 6 classes: Angry, bored, happy, neutral, sad and surprised. Features for an input utterance are extracted from statistics of phonetic and prosodic information. Phonetic information includes log energy, shimmer, formant frequencies, and Teager energy; Prosodic information includes Pitch, jitter, duration, and rate of speech. Finally a pattern classifier based on Gaussian support vector machines decides the emotion class of the utterance. We record speech commands and dialogs uttered at 2m away from microphones in 5 different directions. Experimental results show that the proposed method yields $48\%$ classification accuracy while human classifiers give $71\%$ accuracy.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2005년도 추계 학술대회 발표논문집
• /
• pp.163-166
• /
• 2005

We propose a speech emotion recognition method for natural human-robot interface. In the proposed method, emotion is classified into 6 classes: Angry, bored, happy, neutral, sad and surprised. Features for an input utterance are extracted from statistics of phonetic and prosodic information. Phonetic information includes log energy, shimmer, formant frequencies, and Teager energy; Prosodic information includes pitch, jitter, duration, and rate of speech. Finally a patten classifier based on Gaussian support vector machines decides the emotion class of the utterance. We record speech commands and dialogs uttered at 2m away from microphones in 5different directions. Experimental results show that the proposed method yields 59% classification accuracy while human classifiers give about 50%accuracy, which confirms that the proposed method achieves performance comparable to a human.

• 말소리와 음성과학
• /
• 제12권3호
• /
• pp.25-32
• /
• 2020

본 연구는 영어 발화에 나타난 7가지 감정들(calm, happy, sad, angry, fearful, disgust, surprised)을 분석하고자 감정발화(672개)와 감정중립 발화(48개)와의 운율적 거리를 측정하였다. 이를 위해 외국어 발음평가에 사용되었던 방법을 적용하여 음의 높낮이(Hz), 음의 강도(dB), 음의 길이(sec)와 같은 운율의 3요소를 유클리디언 거리로 계산하였는데 기존연구에서 더 나아가 유클리디언 거리계산 정규화 방법, z-score 방법 그리고 z-score 정규화 방법을 추가해 총 4가지 그룹(sqrF0, sqrINT, sqrDUR; norsqrF0, norsqrINT, norsqrDUR; sqrzF0, sqrzINT, sqrzDUR; norsqrzF0, norsqrzINT, norsqrzDUR)의 방법을 분석에 사용하였다. 그 결과 인지적 측면과 음향적 측면의 분석 모두에서 유클리디언 운율거리를 정규화한 norsqrF0, norsqrINT, norsqrDUR이 일관성 있게 가장 효과적인 측정방법으로 나타났다. 유클리디언 거리계산 정규화 방법으로 감정발화와 감정중립 발화를 비교했을 때, 전반적으로 감정에 따른 운율의 변화는 음의 높낮이(Hz)가 가장 크고 그다음 음의 길이(sec), 그리고 음의 강도(dB)가 가장 작게 나타났다. Tukey 사후검증 결과 norsqrF0의 경우 calm

• 음성과학
• /
• 제12권4호
• /
• pp.113-128
• /
• 2005

The underlying premise of this study was that L1 production is easily transferred into L2 production. In neutral intonation, there is a consonant-tone correlation in Korean: High tone patterns are correlated with voiceless aspirated and tense consonants and Low-High tone patterns are correlated with lax or other voiced consonants. The purpose of this study was to see whether the correlation in Korean (L1) is transferred into English (L2) production and whether the degree of transfer differs depending on the degree of proficiency. Eight Korean speakers and two American speakers participated in the experiment. F0 contours of words and sentences were collected and analyzed. The results of the present study showed that there is a strong correlation between voicing and tone in L2 utterances. When utterance-initial consonant types were voiceless, the word or the sentence began with the H pattern; otherwise it had the LH pattern. The degree of interference differed depending on the degree of proficiency: less proficient speakers showed a stronger correlation in terms of the magnitude (Hz) and size (ms) of the effects on F0. The results indicate that the consonant-tone correlation in L1 is strongly transferred into L2 production and the correlation transfer can be one of the actual aspects that cause L2 speakers to produce deviant L2 accents and intonation.