• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.7
• /
• pp.1380-1386
• /
• 2003

Speech recognition performance is degraded by the variation in vocal tract length among speakers. In this paper, we have used a vocal tract length normalization method wherein the frequency axis of the short-time spectrum associated with a speaker's speech is scaled to minimize the effects of speaker's vocal tract length on the speech recognition performance In order to normalize vocal tract length, we tried several frequency warping functions such as linear and piece-wise linear function. Variable interval piece-wise linear warping function is proposed to effectively model the variation of frequency axis scale due to the large variation of vocal tract length. Experimental results on TIDIGITS connected digits showed the dramatic reduction of word error rates from 2.15% to 0.53% by the proposed vocal tract normalization.

• Journal of the Korean Society of Laryngology, Phoniatrics and Logopedics
• /
• v.17 no.1
• /
• pp.14-16
• /
• 2006

Background and Objectives: The purpose of this study is to investigate the change of vocal tract length according to the level of the pitch by the singers. Materials and Methods: Fifteen tenors were asked to produce successive /a/ sound in G4(382Hz) for the head register, C3(131Hz) for the chest register and usual speaking sound. The control group consisted of 15 males of an similar age who are not professional singers. The length of vocal tract was calculated by applying the formula of Fn=(2n-1) c/4L(F : formant frequency, c : the speed of sound in the vocal tract(350m/sec), L : length of vocal tract, $n=1,2,3,4,{\ldots}{\infty}$). Results: In singer's group, there showed no significant statistical difference of length among head and chest register and usual speaking sound. However in the control group, there showed statistically significant difference of length. Comparison of the absolute difference in the length of vocal tract by changing level of pitch in phonation, between the control group and the singers group. Changing from G4 phonation to C3 phonation and C3 phonation to usual speaking sound showed statistically difference of vocal tract length was less in the singers group than the control group. Conclusion: The change of vocal tract length, in either speaking or singing, was less in singers than the control group. We could assume that the singers maintain their larynx position constantly throughout the pitch range when phonation.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.59 no.6
• /
• pp.1126-1130
• /
• 2010

Speech production can be viewed as a filtering operation in which a sound source excites a vocal tract filter. The vocal tract is modeled as a chain of cylinders of varying cross-sectional area in linear prediction acoustic tube modeling. In this modeling the most common implementation assumes equal length of tube sections. Therefore, to model complex vocal tract shapes, a large number of tube sections are needed. This paper proposes a new vocal tract model with unfixed sectionlengths, which uses the reduced lattice filter for modeling the vocal tract. This model transforms the lattice filter to reduced structure and the Burg algorithm to modified version. When the conventional and the proposed models are implemented with the same order of linear prediction analysis, the proposed model can produce more accurate results than the conventional one. To implement a system within similar accuracy level, it may be possible to reduce the stages of the lattice filter structure. The proposed model produces the more similar vocal tract shape than the conventional one.

• Journal of the Korean Society of Laryngology, Phoniatrics and Logopedics
• /
• v.18 no.1
• /
• pp.56-61
• /
• 2007

Introduction: Actually classification of classic singers' voice depends on habitual judgment by voice teachers or voice trainer referring to vocal timbre, vocal range and vocal quality. Such judgments, however, may turn out to be incorrect because they are based on subjective opinions. Therefore, more objective methodology is required. Method: Foreign dissertations searched through Pub Med, along with foreign and domestic journals, were reviewed regard ing how singers' voice has been categorized. Results: Vocal range, vocal timbre, voice quality, fundamental frequency of habitual speaking, length of vocal tract, the length from cricoid cartilage to thyroid cartilage's thyroid notch and length of vocal fold, tone of passaggio as well as traditional approaches such as perceptual judgment used by professional singers have been used for categorize the voice classification. Conclusion: To optimize categorizing singers' voice, vocal range, vocal timbre, voice quality, fundamental frequency of habitual speaking, length of vocal tract, the length from cricoid cartilage to thyroid cartilage's thyroid notch and length of vocal fold, tone of passaggio may be totally recommended.

• Phonetics and Speech Sciences
• /
• v.9 no.1
• /
• pp.41-47
• /
• 2017

The main advantage of the statistical parametric speech synthesis is its flexibility in changing voice characteristics. A personalized text-to-speech(TTS) system can be implemented by combining a speech synthesis system and a voice transformation system, and it is widely used in many application areas. It is known that the fundamental frequency and the spectral envelope of speech signal can be independently modified to convert the voice characteristics. Also it is important to maintain naturalness of the transformed speech. In this paper, a speech synthesis system based on Hidden Markov Model(HMM-based speech synthesis, HTS) using the STRAIGHT vocoder is constructed and voice transformation is conducted by modifying the fundamental frequency and spectral envelope. The fundamental frequency is transformed in a scaling method, and the spectral envelope is transformed through frequency warping method to control the speaker's vocal tract length. In particular, this study proposes a voice transformation method using the correlation between fundamental frequency and vocal tract length. Subjective evaluations were conducted to assess preference and mean opinion scores(MOS) for naturalness of synthetic speech. Experimental results showed that the proposed voice transformation method achieved higher preference than baseline systems while maintaining the naturalness of the speech quality.

• Journal of radiological science and technology
• /
• v.20 no.1
• /
• pp.21-27
• /
• 1997

Total laryngectomy has become an usual treatment for any advanced carcinoma of the laynx, but most patients who have undergone total laryngectomy have shown permanant disability in voice production. I compared the first three formant frequencies estimated from MRI to those measured directly from speech data of the T-E patients and the normal. It was to estimate the accuracy of MRI and to compare the vocal tract shape of the normal to T-E patients. The obtained results were as follows : 1. The middle sagittle section of the MRI represents vocal tract well during pnonation. The vocal tract shape of the T-E shunt patients are lack of pharyngeal space and superior space of the glottis. 2. The length of the normal subject's vocal tract is 17 cm. For the T-E shunt patients, the length from lip to shunt opening is 17.5 cm in case 1, and 18.5 cm in case 2. That of the true resonante chamber is 13 cm and 13.5 cm for each case respectively. 3. T-E shunt patients phonated strained voice. The intensity of the higher formant frequency decreased especially in /o/, /u/. 4. The vocal tract is shortened during the phonation by T-E shunt patients. In case of /e/ and /i/, front cavities are constricted while back cavities are shortened. 5. The pseudoglottis of the T-E shunt patients is located at $14{\sim}15\;cm$ below from lips.

• Journal of the Korean Society of Laryngology, Phoniatrics and Logopedics
• /
• v.15 no.2
• /
• pp.98-111
• /
• 2004

Background and Objectives : The purpose of this study is to analyze the difference of Fundamental Frequency(Hz), Closed Quotient(Qx ; %), Intensity(dB), Vocal tract length and width(cm), formant frequency(Hz), level of formant frequency(dB) depending on the larynx position. Materials and Methods : One professional male singer(career : 28 years) produced sustained vowel /a/,/e/,/i/,/o/,/u/ in two larynx position (higher, lower) with Dr. Speech and video fluoroscopy was used to quantify the vocal tract morphology. Results : In lower larynx position, CQ is increased 9.8% and Intensity is increased about 10% and level of Formant Frequency is increased. And also Vocal tract length is longer 2.4cm, Vocal tract width(Anterior width : 0.4cm, lateral width : 0.2cm) is wider than in higher larynx position. Conclusions : Singer's formant has a prominent spectrum envelope peak near 2400-2600Hz by clustering of F3, F4 and F5 near 3400Hz in lower larynx position.

• Proceedings of the KIEE Conference
• /
• 2002.07d
• /
• pp.2454-2456
• /
• 2002

Since articuratory synthesis models the human vocal organs as precise as possible, it is potentially the most desirable method to produce various words and languages. This paper proposes a new type of an articulatory synthesizer using Mermelstein vocal tract model and Kelly-Lochbaum digital filter. Previous researches have assumed that the length of the vocal tract or the number of its cross sections dose not vary while uttering. However, the continuous utterance can not be easily implemented under this assumption. The limitation is overcomed by "Buffer Rearrangement" for dynamic vocal tract in this paper.

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.12
• /
• pp.1137-1142
• /
• 2010

This paper studied the speech parameters less affected by the human emotion for the development of the robust speech recognition system. For this purpose, the effect of emotion on the speech recognition system and robust speech parameters of speech recognition system were studied using speech database containing various emotions. In this study, mel-cepstral coefficient, delta-cepstral coefficient, RASTA mel-cepstral coefficient and frequency warped mel-cepstral coefficient were used as feature parameters. And CMS (Cepstral Mean Subtraction) method were used as a signal bias removal technique. Experimental results showed that the HMM based speaker independent word recognizer using vocal tract length normalized mel-cepstral coefficient, its derivatives and CMS as a signal bias removal showed the best performance of 0.78% word error rate. This corresponds to about a 50% word error reduction as compare to the performance of baseline system using mel-cepstral coefficient, its derivatives and CMS.

• Phonetics and Speech Sciences
• /
• v.9 no.2
• /
• pp.85-94
• /
• 2017

Conventional speaker verification systems using time delay neural network, identity vector and probabilistic linear discriminant analysis (TDNN-Ivector-PLDA) are known to be very effective for verifying long-duration speech utterances. However, when test utterances are of short duration, duration mismatch between enrollment and test utterances significantly degrades the performance of TDNN-Ivector-PLDA systems. To compensate for the I-vector mismatch between long and short utterances, this paper proposes to use probabilistic linear discriminant analysis (PLDA) model adaptation with augmented data. A PLDA model is trained on vast amount of speech data, most of which have long duration. Then, the PLDA model is adapted with the I-vectors obtained from short-utterance data which are augmented by using vocal tract length perturbation (VTLP). In computer experiments using the NIST SRE 2008 database, the proposed method is shown to achieve significantly better performance than the conventional TDNN-Ivector-PLDA systems when there exists duration mismatch between enrollment and test utterances.