Feature extraction using Grassmann representation and its structural metric for brain signal processing
| Project/Area Number |
24700163
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | The University of Electro-Communications |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2012-04-01 – 2015-03-31
|
| Project Status |
Completed (Fiscal Year 2014)
|
| Budget Amount *help |
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2014: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2013: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2012: ¥3,120,000 (Direct Cost: ¥2,400,000、Indirect Cost: ¥720,000)
|
|---|
| Keywords | パターン識別 / 機械学習 / 生体信号処理 / 脳信号処理 / グラスマン多様体 / カーネルトリック / 脳コンピュータインターフェース / 脳波 / EEG / カーネル主成分分析 / パターン認識 / 事象関連脱同期 / Grassmann多様体 / Mahalanobis距離 / 共空間パターン |
|---|
| Outline of Final Research Achievements |
We treat problems to extract intrinsic features efficiently from variational patterns, such that patterns are obtained as sequential images or faces, time series of brain signals. We approximate these patterns by linear subspaces, and investigate the structure of the set of linear spaces that is Grassmann manifold or Grassmannian. We introduced a probabilistic distance metric, Mahalanobis distance to Grassmannian to improve the classification performance. Furthermore, we extend the Grassmannian representation by using the soft-thresholding technique. The standard Grassmannian representation hardly divide the signal subspace and noise subspace, on the other hand, the extended Grassmannian representation softly divides them. We applied these proposed techniques to various classification problems including brain signal processing, and show its performance.
|
Report
(4 results)
Research Products
(12 results)
