| Presentation | 2006/9/23 Face recognition using feature fusion and information criterion : K-nearest neighbor learning based on Mahalanobis distance Takami SATONAKA, Keiichi UCHIMURA, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | We describe the adaptive metric learning procedure combining the asymptotic variance estimation and fine adjustment of the metric parameters for the face recognition. The metric learning model based on the Maha-lanobis distance suffered from the degraded performance due to the limitation of available training samples. For accurate estimation of metric parameters, the asymptotic model sharing common variance among classes with difference averages has been introduced. The MDL criterion is formulated to decide the optimal number of classes in the asymptotic statistical model. The feature fusion method assuming the synthesized feature patterns is proposed to construct the Kth-nearest neighbor learning network employing multiple Mahalanobis decision function and improves the performance with the increase of the metric function per class. We compare the simulation results using the ORL databases with these of other systems. The hybrid of the asymptotic variance estimation and fine adjustment of the metric parameters has attained the minimal recognition error rate. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | face recognition / Mahalanobis distance / Information criterion / MDL |
| Paper # | HCS2006-48 |
| Date of Issue |
| Conference Information | |
| Committee | HCS |
|---|---|
| Conference Date | 2006/9/23(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Human Communication Science (HCS) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Face recognition using feature fusion and information criterion : K-nearest neighbor learning based on Mahalanobis distance |
| Sub Title (in English) | |
| Keyword(1) | face recognition |
| Keyword(2) | Mahalanobis distance |
| Keyword(3) | Information criterion |
| Keyword(4) | MDL |
| 1st Author's Name | Takami SATONAKA |
| 1st Author's Affiliation | Kumamoto Prefectual College of Technology() |
| 2nd Author's Name | Keiichi UCHIMURA |
| 2nd Author's Affiliation | Faculty of Engineering, Kumamoto University |
| Date | 2006/9/23 |
| Paper # | HCS2006-48 |
| Volume (vol) | vol.106 |
| Number (no) | 268 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |