compariSong Files

CompariSONG : Query-by-humming melody search engine
 
Step 0.
Check whether you have Java runtime environment (JRE). Chances that you have it are high. 
Open a DOS prompt (a.k.a command window) or Linux shell and enter the following command
java -version
The output will look something like:
java version "1.6.0_17"
Java(TM) SE Runtime Environment (build 1.6.0_17-b04)
Java HotSpot(TM) Client VM (build 14.3-b01, mixed mode, sharing)
Go to Step 2. 

If this results in an error message like the following, your computer has no version of Java installed:
'java' is not recognized as an internal or external command, operable program or batch file.
Then go to step 1.

Step 1. 
Install Java SE Runtime Environment (JRE) 6.0 or higher -> jre-6u3-windows-i586-p-s.exe 
or 
download latest version at http://java.sun.com/javase/downloads/index.jsp




Step 2. 
Unzip the contents of comparisong.zip 

Step 3.
 Browse to the unzipped folder

Step 4. 
If on win32 start run.bat
If on win64 start run64.bat
If on Linux start run.sh
Mac OS X version could be provided on demand. 

Usage: 
Open a database through File-> Open database. 
Open a query through File-> Open query. 
You can use test database folder which is an example test database directory. Then load a query. In test database/query are some already hummed queries. 
Search. The results are listed based on similarity.
After selesting one of the results you can play it to check whether this is the intended song. 
Enjoy!


Creating own dababase and queries 
(NOTE: if u cannot sing, just use the "test database" directory from comparisong.zip  )

1. make sure files/features.xml and files/settings.xml and 
icons/comparisong.png
are in current dir.

2. create a directory <db> for database. 
Put *.mp3 files inside. These will be the actual search musical data. 
Add subdirectory called "tagged" and inside it put sung tagged entries in .wav format.
These are counterparts of real .mp3 files sung by user before-hand in .wav format by a database administarator.
So in order for the system to work, each target .mp3 song should has a tagged entry, because system actually compares query with tagged entires.
This is needed because we cannot extract robustly only voice from real .mp3 files :(
 

NOTE: the name of tagged songs should be <artist> - <song>.wav.
 Each tagged entry should have the real matching .mp3 in the <db directory>





For any comments write to Georgi Dzhambazov at joro.dzhambazov@gmail.com