| Project/Area Number |
15K16395
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Rehabilitation science/Welfare engineering
|
|---|
| Research Institution | Nara Institute of Science and Technology |
|---|
Principal Investigator |
Kubo Takatomi 奈良先端科学技術大学院大学, 情報科学研究科, 特任准教授 (20631550)
|
|---|
| Project Period (FY) |
2015-04-01 – 2018-03-31
|
| Project Status |
Completed (Fiscal Year 2017)
|
| Budget Amount *help |
¥3,250,000 (Direct Cost: ¥2,500,000、Indirect Cost: ¥750,000)
Fiscal Year 2017: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000)
Fiscal Year 2016: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2015: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
|
|---|
| Keywords | 時系列分節化 / ノンパラメトリックベイズ法 / 深層学習 / 表面筋電位信号 / 分節化 / 多点表面筋電位信号 / 機会学習 / 発話 |
|---|
| Outline of Final Research Achievements |
When deep learning is applied to uncommon data like surface electromyographic signals during a speech, it is necessary to determine parameters, e.g., the number of layers in the network, the number of nodes in each layer, etc. appropriately. In 2015, we developed a method to determine them depending on data, and reported it in a domestic conference, international conference, and a journal. In 2017, we developed a method that can segment time-series data with the unknown number of states by using two Bayesian non-parametric methods, Beta process autoregressive hidden Markov model and Hierarchical Pitman- Yor language model. We reported it in a domestic conference and an international conference. We received a recommendation to a special issue of a journal for our presentation in the domestic conference.
|
