| Budget Amount *help |
¥2,200,000 (Direct Cost: ¥2,200,000)
Fiscal Year 2006: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 2005: ¥1,500,000 (Direct Cost: ¥1,500,000)
|
|---|
| Research Abstract |
This research aimed at the development of direct and intuitive operational walking guide tactile display device (TDG : Tactile Display Glove) with the keyword of real-time walking rapid achievement, surrounding state recognition not only but also accurate distance perception to the barriers. We made essential trials on the analysis about the recognition characteristics of tactile stimulator allocation pattern and stimulating method (variable pressure method, variable stroke, variable stimulating frequency method, variable point stimulating method-distance range mapping) From this investigation, the last distance range based variable point stimulating method shows the best and extremely high performance in repetitive distance recognition accuracy on the root of finger on backhand, however, the other method only show the low recognition rate performance. To improve the resolution of the sensitivity (: the resolution of distance recognition) and the recognition rate correspond to average
… More
walking speed, we investigated other stimulator allocation part and pattern using forearm outer side.
For the surrounding state transfer, we developed the stereo vision system to extract and recognize surround state not only the distance map. We proposed a mutual coupling neural network model with an extended a Hebb learning scheme to improve the state identification accuracy improvement. These analysis and the performance are presented in international conferences and published in the book of 'Mobile Robots : Perception ＆ Navigation' from Advanced Robotics Systems International.
The brief result are listed as follows :
1. Development of the direct and intuitive operation tactile display glove ( 5 types )
Distance recognition comparison among several stimulator allocation ( at finger roots, forearm outside (linear : high density, low density, double parallel allocation, digit mode display). Linear allocation on forearm shows the rapid recognition and 2 points discriminative limit improvement)
2. Development of the 3-D distance map calculation algorithm for stereo camera vision
3. Development of the surrounding state and target recognition scheme with extended Hebb learning method
High incrementation average identification accuracy : 30%(simple Hebb learning)→ 72%
4. A proposal of surrounding space perception check method using space recognition sheet
(Constructed/recognized surrounding distance map visualization)
5. Investigation of fast and accurate recognition performance of assign mode and linear allocation mode concominant use TDG
6 Discloser of research on International conferences and Publication (presentation on domestic conference with students) Less
|
