- Youtube demo Here.
- Robust Audio Recognition
- High Accuracy: 100% Accuracycy of 500 6-second songs cuts in 100 songs list.
- High efficiency: recognize in less than 0.1 second per song.
- Memory reduction: 4G is enough for 10000 songs.
- Data Structure: support 2,500,000 songs and each song less than 7 minutes.
- More feature: return the TIME POINT of the cut song in original song.
- fftw3:
If under Debian/Ubuntu, run
apt-get install libfftw3-dev
.
- extract songs to ./bin/ dir.
make
to produce two excutive files in ./bin/- The songs should be .wav format.
- The songs should be contained in ./bin/samples/
- You may download the songs. link: http://pan.baidu.com/s/1ntsoIlr password: wgi7
- run
cd bin
,./build
and it'll produce a file calleddatabase
- run
./recog ${filename}
to recog the songs that in samples list.
- The code is too dirty written by a first-year c++ programmer(me).. :(
- More Documents and comments about the codes.
- All the codes in Windows.
- Transfer the codes from windows to linux.
- Solve the BaiduMusic download tools to update new songs automatically.