| 摘要 |
By using the wavelet method and the inverse Mallat transform of the wavelet theory, frequency analysis of large scale time series data is carried out at high speed while effectively utilizing the hardware of a parallel computer of distributed memory type. To be concrete, on the parallel computer of distributed memory type, data to be analyzed, a scaling function, and a Daubechies's progression are inputted to individual memories. Highest order frequency component parallel numerical integral operation is conducted to derive a frequency component of the highest order to be analyzed. Series low frequency component analyzing using the inverse Mallat transform is conducted to effect frequency analysis other than the highest frequency by using the inverse Mallat transform of the wavelet theory. Frequency data combining is conducted to put together frequencies derived by respective processors in an external memory./!
|
