| Project/Area Number |
17300182
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Rehabilitation science/Welfare engineering
|
|---|
| Research Institution | Japan Advanced Institute of Science and Technology |
|---|
Principal Investigator |
DANG Jianwu Japan Advanced Institute of Science and Technology, School of Information Science, Professor (80334796)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HONDA Kiyoshi ATR, Researcher (90395088)
AKAGI Masato JAIST, School of Information Science, Professor (20242571)
SUZUKI Noriko Riowa University, Dental School, Associate Professor (10112731)
LU Xugang JAIST, School of Information Science, Assistant Professor (20362022)
|
|---|
| Project Period (FY) |
2005 – 2007
|
| Project Status |
Completed (Fiscal Year 2007)
|
| Budget Amount *help |
¥10,000,000 (Direct Cost: ¥9,400,000、Indirect Cost: ¥600,000)
Fiscal Year 2007: ¥2,600,000 (Direct Cost: ¥2,000,000、Indirect Cost: ¥600,000)
Fiscal Year 2006: ¥2,800,000 (Direct Cost: ¥2,800,000)
Fiscal Year 2005: ¥4,600,000 (Direct Cost: ¥4,600,000)
|
|---|
| Keywords | Rehabilitation / Medicine / Welfare / Information Technology / Physiological computation model / Assisting pronounce training / 舌部分切除 / 発話障害 / 発話状態の推定 / MRI計測 / 生理学的モデル / 音声合成・音声処理 / 有限要素法 / 発話訓練 |
|---|
| Research Abstract |
Organic defects of the speech organs and inborn/acquired profound hearing impairments would cause some serious problems in speech intelligibilities. One of the organic defects results from the glossectomy that impairs the morphology and functions of the speech organs. However, it is difficult to predict speech dysfunctions caused by the glossectomy and to build a plan for rehabilitations after the glossectomy. This study aims to minimizes the problems by means of predicting the potential impairments caused by the glossectomy and assisting speech training using a physiological articulatory model.
To predict the impairments of speech after the glossectomy, we extended the partial 3D physiological tongue model, which was previously constructed at our Lab, to a full 3D model, and designed a muscle organization that can be controlled asymmetrically. As a clinical application, we followed up one patient who has a tumor in his tongue. We adapted MRI based observations of the tumor to the physi
… More
ological model and predicted possible situations resulting from the glossectomy before the surgery, and then examined the simulations after the surgery. The results showed that the deviation of the tongue in protrusion movements and the difference in left and right bending movements by the simulations were consistent with those of the observation. This study showed an example that bridges the fundamental research and clinical applications.
To construct an assisting system for speech training, we developed a method for inverse estimation from speech sound to the vocal tract shape using the physiological articulatory model. In this study, we focused on two problems: how to reduce the one-to-many problem and how to combine coarticulation in the articulatory model based speech synthesizer. To reduce the one-to-many problem, we introduced a novel constraint, in which we separate the possible vocal tract shapes for producing vowels into normal articulations and unusual articulations such as those used by ventriloquists, and then excluded the unusual articulation from the candidate solutions of the inverse estimation.
To model coarticulations in speech, we proposed a carrier model of coarticulation, in which the articulatory movements are treated as that the vocalic movement (the carrier wave) is modulated by the consonantal movement (a modulation wave). The parameters of the carrier model were optimized by observations obtained by the electromagnetic articulographic system. Alter implementing the learned carrier model to the physiological model based speech synthesizer, the synthetic sounds become more natural than those from the previous one. This implies that the carrier model of coarticulation can describe the speech planning of human speech production.
One of the important topics for speech synthesis is to generate individual properties. In the study, we analyzed individual characteristics of the speech and applied to the results to speaker identification. Less
|
