| Project/Area Number |
08650436
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
情報通信工学
|
|---|
| Research Institution | KYUSHU INSTITUTE OF TECHNOLOGY |
|---|
Principal Investigator |
MATSUOKA Kiyotoshi KYUSHU INSTITUTE OF TECHNOLOGY Engineering, Professor, 工学部, 教授 (90110840)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TOKUNARI Tsuyoshi KYUSHU INSTITUTE OF TECHNOLOGY Center for Cooperative Research, Assestant, 地域共同研究センター, 助手 (00237075)
|
|---|
| Project Period (FY) |
1996 – 1997
|
| Project Status |
Completed (Fiscal Year 1997)
|
| Budget Amount *help |
¥1,200,000 (Direct Cost: ¥1,200,000)
Fiscal Year 1997: ¥500,000 (Direct Cost: ¥500,000)
Fiscal Year 1996: ¥700,000 (Direct Cost: ¥700,000)
|
|---|
| Keywords | blind signal separation / blind equalizer / independent component panalysis / high-order statistics / convolutive mixture / non-Gaussianness / ブラインド等価器 / 独立信号分離 / QAM / 非ガウス |
|---|
| Research Abstract |
Sensor fusion is a new trend of measurement technology. Its strategy is to extract as much information as possible from an object consisting of a number of signal sources, using a lot of sensing devices of possibly different modalities. IN general, each sensor's output is a mixture of some source signals. If the transfer function that couples the sources and sensors is know, then the source signals can be recovered by applying its inverse to the sensor signals. However it is not easy to find the transfer function analytically as well as experimentally. In this case, recently a new technique that is called Blind Signal Separation has received a great attention. The technique uses only a priori knowledge (the fact that source signals are mutually statistically independent) to estimate source signals from sensor signals.
We presented a generalization of the Godard algorithm for blind equalization of the communication channels. We have shown that our methods possessed the capability of phase tracking. In the case that the measured signals contain mixtures of both sub-Gaussian (with negative kurtosis) and super-Gaussian (with positive kurtosis) sources, conventional methods do not work well. Therefore we proposed a new approach, in which a set of evaluation functions are introduced and they are minimized one by one. Moreover, we proposed a new algorithm using Newton method, which provided high-speed convergence.
|
