Hung‐Yan Gu

National Taiwan University of Science and Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Chyi‐Yeu Lin

National Taiwan University of Science and Technology

Chin‐Shyurng Fahn

National Taiwan University of Science and Technology

Kuo‐Liang Chung

National Taiwan University of Science and Technology

Chiu-yu Tseng

Institute of Linguistics, Academia Sinica

Jia-Wei Chang

First Affiliated Hospital of Anhui Medical University

Sung-Fung Tsai

National Taiwan University of Science and Technology

Kai‐Jay Lu

National Taiwan University of Science and Technology

Lin-Shan Lee

National Taiwan University

Sung-Feng Tsai

National Taiwan University of Science and Technology

Ming-Yen Lai

National Taiwan University of Science and Technology

Cooperative Institutions

National Taiwan University of Science and Technology

National Taiwan University

National Yang Ming Chiao Tung University

National Cheng Kung University

National Taipei University of Technology

Institute of Information Science, Academia Sinica

Academia Sinica

Chang Gung Memorial Hospital

National Central University

National Chung Hsing University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Mandarin syllable signal synthesis using an HNM based scheme

International Conference on Audio, Language and Image Processing (2008)

Hung‐Yan Gu Yen-Zuo Zhou

In this paper, HNM (harmonic plus noise model) is enhanced and used to design a scheme for synthesizing Mandarin syllable signals. Each syllable is recorded once only and used to synthesize syllable signals with diverse prosodic characteristics without suffering significant signal-quality degradation. For a control point on the synthetic syllable's time axis, two corresponding analysis frames' HNM parameters are interpolated to derive the HNM parameters for the control point. Furthermore, for pitch-contour tuning, another timbre-reserving interpolation is performed for the HNM parameters on a control point. Then, signal samples are synthesized with the HNM synthesis equations rewritten here. According to the result of the perception tests, the HNM based scheme proposed here can indeed be used to synthesize syllable signals with consistent timbre and high signal clarity.

Mandarin Chinese

Timbre

SIGNAL (programming language)

10.1109/icalip.2008.4590153

Cite

Citations (3)

Improving segmental GMM based voice conversion method with target frame selection

Hung‐Yan Gu Sung-Fung Tsai

In this paper, the voice conversion method based on segmental Gaussian mixture models (GMMs) is further improved by adding the module of target frame selection (TFS). Segmental GMMs are meant to replace a single GMM of a large number of mixture components with several voice-content specific GMMs each consisting of much fewer mixture components. In addition, TFS is used to find a frame, of spectral features near to the mapped feature vector, from the target-speaker frame pool corresponding to the segment class as the input frame belongs to. Both ideas are intended to alleviate the problem that the converted spectral envelopes are often over smoothed. To evaluate the performance of the two ideas mentioned, three voice conversion systems are constructed, and used to conduct listening tests. The results of the tests show that the system using the two ideas together can obtain much improved voice quality. In addition, the measured variance ratio (VR) values show that the system with the two ideas adopted also obtains the highest VR value.

Feature (linguistics)

10.1109/iscslp.2014.6936633

Cite

Citations (1)

以線性多變量迴歸來對映分段後音框之語音轉換方法 (A Voice Conversion Method Mapping Segmented Frames with Linear Multivariate Regression) [In Chinese].

International Conference on Computational Linguistics (2012)

Hung‐Yan Gu Jia-Wei Chang Zan-Wei Wang

Source

Cite

Citations (0)

Integrating speaker-nonspecific timbre transformation to an HNM-based speech synthesis scheme

Journal of the Chinese Institute of Engineers (2012)

Hung‐Yan Gu Chen-Lin Tsai

Abstract In this article, a speech signal synthesis scheme based on harmonic-plus-noise model (HNM) is extended to provide the function of speaker-nonspecific timbre transformation. To transform synthetic speech's timbre, we have developed a formant-based frequency mapping method called piece-wise linear frequency mapping (PLFM). Additionally, a commonly adopted method is frequency axis scaling (FAS). Both methods have been integrated into the HNM-based signal synthesis scheme, and a real-time speech synthesis system has been implemented according to this scheme. Using the speech files synthesized, perception tests are conducted. The results show that the proposed scheme can indeed transform the source timbre of a female adult into the timbre of a male adult, boy, or girl. In addition, the method PLFM is shown to be better than FAS in obtaining more masculine timbre. Keywords: speech synthesistimbre transformationharmonic plus noise modelfrequency axis scaling Acknowledgments The financial support provided by National Science Council, Taiwan, under the grant, NSC 98–2221–E–011–116, is gratefully acknowledged.

Timbre

Multidimensional scaling

SIGNAL (programming language)

10.1080/02533839.2012.730267

Cite

Citations (1)

Model spectrum-progression with DTW and ANN for speech synthesis

Hung‐Yan Gu Chang-Yi Wu

In this paper, an ANN based spectrum-progression model (SPM) is proposed. This model is intended to improve the fluency level of synthetic Mandarin speech under the situation that only a small training corpus is available. In constructing this model, first each target syllable is matched with its reference syllable by using DTW. Then, each warped path, i.e. spectrum-progression path, is time normalized to fixed dimensions, and used to train an ANN based SPM. After training, the SPM is used together with other modules such as text analysis, prosody parameter generation, and signal sample generation to synthesize Mandarin speech. Then, the synthetic speech is used to conduct perception tests. The test results show that the SPM proposed here can indeed improve the fluency level noticeably.

Mandarin Chinese

10.1109/ecticon.2009.5137216

Cite

Citations (7)

A new speech enhancing scheme combining NLMS, fuzzy logic and Kalman filtering

Maolin Chen Hung‐Yan Gu Ching‐Long Shih

This paper presents an improvement to ordinary voice filters that are unable to remove low frequency noise. This problem occurs because their filter parameters are over tuned, thereby destroying the speech characteristics. The proposed scheme combines NLMS, Fuzzy logic and Kalman filtering to restrain the background noise and keep the speech characteristics. This scheme is called the Normalized Fuzzy Logic Kalman Filter (NFLKF). It is especially effective when speech signals are collected in a noisy environment. Here, the output signal of the Kalman filtering is analyzed with the normalized LMS to obtain the coefficient, sigma _k , and is also analyzed with the fuzzy logic to obtain the threshold, f _k . Then, sigma _k and f _k are used together to adjust the Kalman filter parameters. This scheme can restrain the noise and improve the signal-to-noise ratio. The empirical validation is done by comparing the spectrogram from our scheme with the spectrograms from other filtering schemes, including the Normalized Least Mean Square (NLMS) Filter, the Kalman Filter and the Recursive Least Square Filter (RLS). The results show that the filtering schemes proposed can indeed restrain medium and low frequency noises which are usually difficult to handle, and does not compromise the speech characteristic. Therefore, a better signal-noise ratio is obtained and the speech quality is enhanced.

Spectrogram

10.1109/fuzzy.2008.4630388

Cite

Citations (2)

A large-alphabet-oriented scheme for Chinese and English text compression

Software Practice and Experience (2005)

Hung‐Yan Gu

In this paper, a large-alphabet-oriented scheme is proposed for both Chinese and English text compression. Our scheme parses Chinese text with the alphabet defined by Big-5 code, and parses English text with some rules designed here. Thus, the alphabet used for English is not a word alphabet. After a token is parsed out from the input text, zero-, first-, and second-order Markov models are used to estimate the occurrence probabilities of this token. Then, the probabilities estimated are blended and accumulated in order to perform arithmetic coding. To implement arithmetic coding under a large alphabet and probability-blending condition, a way to partition count-value range is studied. Our scheme has been programmed and can be executed as a software package. Then, typical Chinese and English text files are compressed to study the influences of alphabet size and prediction order. On average, our compression scheme can reduce a text file's size to 33.9% for Chinese and to 23.3% for English text. These rates are comparable with or better than those obtained by popular data compression packages. Copyright © 2005 John Wiley & Sons, Ltd.

Arithmetic coding

10.1002/spe.661

Cite

Citations (1)

基於ANN之頻譜演進模型及其於國語語音合成之應用 (An ANN based Spectrum-progression Model and Its Application to Mandarin Speech Synthesis) [In Chinese]

International Conference on Computational Linguistics (2008)

Hung‐Yan Gu Chang-Yi Wu

Mandarin Chinese

Source

Cite

Citations (0)

An acoustic and articulatory knowledge integrated method for improving synthetic Mandarin speech's fluency

Hung‐Yan Gu Kuo-Hsian Wang

In synthetic Mandarin speech, discontinuity of formant traces at syllable boundaries is a key factor that lowers the fluency level. Therefore, we study an acoustic and articulatory knowledge integrated method to solve this discontinuity problem. First, representative trisyllable contexts are selected and their signals are recorded. The signal of the middle syllable of each trisyllable pronunciation is then extracted to make a synthesis unit. To select a synthesis unit among multiple candidates, a distance function is defined to measure the spectral similarity between two synthesis units to be concatenated. In addition, several linking-restriction rules are derived, according to articulatory knowledge, to prevent some synthesis units being linked into a sequence. Then, a globally best synthesis-unit sequence is searched by using a dynamic programming based algorithm. When this method is applied, the formant traces at syllable boundaries become smoother. Also, subjective evaluation shows that the fluency level of synthetic Mandarin speech can indeed be improved a lot.

Mandarin Chinese

Pronunciation

Discontinuity (linguistics)

10.1109/chinsl.2004.1409622

Cite

Citations (5)

Mandarin singing voice synthesis using ANN vibrato parameter models

International Conference on Machine Learning and Cybernetics (2008)

Hung‐Yan Gu Zheng-Fu Lin

In this paper, the vibrato parameters of sung syllables are analyzed by using short-time Fourier transform and the method of analytic signal. After the vibrato parameter values for all training syllables are obtained, they are used to train an artificial neural network (ANN) for each type of vibrato parameter. Then, these ANN models are used to generate the values of vibrato parameters. Next, these parameter values and other music information are used together to control a harmonic-plus-noise (HNM) model to synthesize singing voice signals. With the synthetic singing voice, subjective perception tests are conducted. The result show that the singing voice synthesized with the ANN generated vibrato parameters is apparently more natural than the singing voice synthesized with fixed vibrato parameters.

Vibrato

10.1109/icmlc.2008.4620973

Cite

Citations (1)