Multi-speaker articulatory trajectory formation based on speaker-independent articulatory HMMs

Sano, Hiroya; Mochida, Takemi

doi:10.1016/j.specom.2006.08.002

Cited by 13 publications

(19 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The estimation error of articulatory parameters for NTD (0.16 mm) was much smaller than 1.22 mm for an articulatory HMM [13] and 1.65 mm for a kinematic triphone model [14] because NTD parameters have not been statistically modeled in this study. But we expect that the error of NTD with statistical modeling will be equivalent to or smaller than that of other articulatory modeling by using redundancy in NTD parameters.…”

Section: Resultsmentioning

confidence: 85%

Non-negative temporal decomposition of speech parameters

Sano¹

2010

2010 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

We present a non-negative temporal decomposition method for line spectrum pair and articulatory parameters. Based on the multiplicative update rules derived from a non-negative matrix factorization algorithm, these parameters decompose into a set of temporally overlapped event functions that are restricted to the range [0, 1] and corresponding event vectors. With the proposed method, the RMS error of the measured and estimated articulatory parameters is 0.16 mm and the spectral distance of the measured and estimated line spectrum pair parameters is 1.97 dB. These results also show that these estimation errors of proposed method are significantly smaller than those of the conventional method. This technique will be useful for many applications, such as speech coding and speech modification.

show abstract

Section: Resultsmentioning

confidence: 85%

Non-negative temporal decomposition of speech parameters

Sano¹

2010

2010 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Multi-speaker acoustic-to-articulatory inversion based on hidden Markov models (HMM) have been used in [17]; however it requires a data stream with information about the phonemes present in the speech signal. Additionally, the training of the HMM models in [17] requires several speakers, whereas only one speaker is required in the method proposed in this paper.…”

Section: Databasementioning

confidence: 99%

“…Additionally, the training of the HMM models in [17] requires several speakers, whereas only one speaker is required in the method proposed in this paper. It is important to note that EMA (ElectroMagnetic Articulograph) data are scarce and expensive.…”

Section: Databasementioning

confidence: 99%

Subject-independent acoustic-to-articulatory mapping of fricative sounds by using vocal tract length normalization

Sepúlveda-Sepúlveda

Castellanos-Domínguez

Gómez-Vilda

2015

Rev. Fac. Ing. Antioquia

View full text Add to dashboard Cite

“…In the speaker-dependent manner, there are methods such as codebook mapping [1], mixture density network [2], Gaussian mixture model (GMM) based mapping [3], and hidden Markov model (HMM) based methods [4][5][6][7]. In the study of the speaker-adaptation for the inversion problem, Hueber et al [8] use an approach that merges the voice conversion step and the acoustic-articulatory inversion step; Hiroya and Honda [9] use a speaker adaptation method in HMMbased speech production model; Hiroya and Mochida [10] proposed a multi-speaker articulatory trajectory formation based on speakerindependent articulatory HMMs using adaptive training.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of linear regression for speaker adaptation in HMM-based articulatory movements estimation

Tao

Wang

2015

2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Acoustic-to-articulatory inversion problem is usually studied in speaker-specific manner because both articulatory data and acoustic features contain speaker-specific components. This paper presents our work on speaker-adaptation training for this problem. We implement speaker adaptation in HMM-based acoustic-to-articulatory inversion mapping, and evaluate different combinatorial structures of the articulatory data and acoustic features. The HMM-based inversion mapping models are built with single-stream and multistream, independent clustering and shared clustering structures. The speaker adaptation is implemented in stream-independent structure and shared adaptation structure. The constrained maximum likelihood linear regression method is used for the speaker-adaptive transformation. The experimental results show that the sharing of the speaker-adaptive transformation of the articulatory feature stream and acoustic feature stream can improve the estimation accuracy in inversion mapping. The multi-stream system with shared clustering and shared adaptive transformation has the best result among all the tested structures.Index Terms-speaker adaptation, acoustic-to-articulatory inversion, maximum likelihood linear regression HMM-BASED ARTICULATORY MODEL Single-stream structureThe single-stream model is the HMM with only a stream of articulatory data. The framework of the HMMs training for articulatory features is shown in Figure 1. Let

show abstract

Multi-speaker articulatory trajectory formation based on speaker-independent articulatory HMMs

Cited by 13 publications

References 26 publications

Non-negative temporal decomposition of speech parameters

Non-negative temporal decomposition of speech parameters

Subject-independent acoustic-to-articulatory mapping of fricative sounds by using vocal tract length normalization

Evaluation of linear regression for speaker adaptation in HMM-based articulatory movements estimation

Contact Info

Product

Resources

About