| Project/Area Number |
03650306
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
情報工学
|
|---|
| Research Institution | Kyushu University |
|---|
Principal Investigator |
SAKOE Hiroaki Kyushu Univ.,Fac. of Eng., Professor, 工学部, 教授 (30215701)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KATAYAMA Yoshinori Kyushu Univ.,Fac. of Eng., Research Associate, 工学部, 助手 (00214338)
MIYAZAKI Akio Kyushu Univ.,Fac. of Eng., Associate Professor, 工学部, 助教授 (70192763)
|
|---|
| Project Period (FY) |
1991 – 1992
|
| Project Status |
Completed (Fiscal Year 1992)
|
| Budget Amount *help |
¥1,900,000 (Direct Cost: ¥1,900,000)
Fiscal Year 1992: ¥300,000 (Direct Cost: ¥300,000)
Fiscal Year 1991: ¥1,600,000 (Direct Cost: ¥1,600,000)
|
|---|
| Keywords | On-line Character Recognition / Stroke Order Free / Stochastic Model / Neural Network / Dynamic Programming / HMM / 2-Dimensional HMM / Neural Prediction Model / 筆順フリ- / 動的計画法 |
|---|
| Research Abstract |
Stroke-order-free online hand-written character recognition algorithms were investigated based on dynamic programming based structural analysis, neural networks, and/or stochastic model.
A new recognition framework was proposed which evaluates likelihood measure between LR (Left-to-Right) type Hidden Markov Model (one-dimensional model) and dot-represented character pattern (two-dimensional pattern). This 1-2 dimensional HMM algorithm achieved stroke-order-free feature by directly accepting two-dimensional input. Learning strategy and 1-2 dimensional likelihood evaluation algorithm were established.
Furthermore, efficiency improvement of the algorithm were investigated. Memory and computation efficiencies were enhanced by order of magnitude compared with the basic algorithm.
Besides the above investigations, applications of neural networks to the online hand-written character recognition problems were explored. Neural prediction model which adaptively predicts pen movement was proposed and successfully evaluated.
|
