|
|
|||||||||||||||||||
![]() |
|||||||||||||||||||
|
The MusicHunt is a project for researching how various sparse representations of music signals can be used to classify music. Little is known about sparse representations and the aim of this project is mainly to learn about properties of these representations. To do this we need to calculate a long series of sparse representations of many music signals. Unfortunately, it requires quite a lot of computational power to do this. However, the good news is that these calculations are easily parallellized. We have done this through a few Matlab routines. We are now asking for your help in performing the actual calculations. To do this, send an email to me, and I'll include you in the list of participants. Note that you cannot participate before I have created a specific file in the computation directory. 1. Map network driveAll you need to participate is a PC with Matlab 6.5 or 7 installed, and mapping of a network drive. The procedure for mapping a network drive is the following (you need to do this only once):
When opening the new drive you see two directories called "disk1" and "disk2" and a screen saver program called MusicHunt.scr (or just MusicHunt) 2a. Setup link for manually starting computations
2b. Setup screen saver for automatically starting computations
3. Doing computationsIf you have setup the computations to start automatically, the screen saver starts matlab whenever a regular screen saver would have kicked in. After spawning a matlab process the computations starts where they left off. The screen saver DOES NOT terminate when you move the mouse or push a key on the keyboard. Rather, you should press the "Terminate Matlab" button in the saver. This flags Matlab to terminate the computations. To ease the load on the processing this flag is only checked between calcuations, so it might take a while (from 1 sec to 2-3 min) before it actually terminates. You can use the "Kill" button, too, if you really need to terminate it fast (but it is not recommended). Only a few minutes of computations are lost this way, however. Note that the spawned matlab process has low priority so I shouldn't affect the "response time" of your computer like a regular, busy matlab session sometimes does. |
|||||||||||||||||||
|
© Copyright 1999-2007 Anders la Cour-Harbo. All rights reserved. This page updated July 9, 2007 |
|||||||||||||||||||