Below are some resulting sounds from the source separation algorithm we have designed, based on a previous melody detection, followed by a spectral Wiener filtering of the original signal. The songs we used are from the ISMIR 2004 Audio Melody Extraction Contest database (see the dedicated ismir 2004 website for more details and to download the test set and the reference files). We suggest you to use headphones in order to be able to hear slight differences and artifacts in the resulting sounds. To go back to the page with the flash player, click here.
The parameters for the analysis of MIREX 2004 database songs are:
- sampling rate: 44100Hz
- length of analysis windows: 46.44 ms (2048 samples)
- hopsize: 5.8 ms (256 samples)
- frequency range for melody detection: Fmin = 100Hz, Fmax = 800Hz (except when precised)
- frequency quantization for the melody: discretized to every 8th tone (that is to say there are 48 frequencies per octave)
- length of the songs: between 14s and 25s
- Title: title name of the song in original test set
- Original: the original song
- Sep. Singer: estimated/separated singer signal
- Sep. Music: estimated/separated music signal
- Remix: left channel = estimated singer, right channel = estimated music
- PitchMatch: percentage of "correctness" in pitch estimation in the song's pitched frames (see the ismir 2004 website for a description)
- TotalMatch: percentage of "correctness" in pitch estimation in the whole song (also taking into account the silences in the singer reference track)
| Title |
Original |
Sep.
Singer |
Sep. Music |
Remix |
PitchMatch (%) | TotalMatch (%) |
| opera_fem2 |
mp3 |
mp3 |
mp3 |
mp3 |
70.2 |
63.7 |
| opera_fem4 |
mp3 | mp3 |
mp3 |
mp3 |
78.8 |
80.2 |
| opera_male3 |
mp3 |
mp3
|
mp3 |
mp3 |
65.5 |
66.8 |
| opera_male5 |
mp3 |
mp3 |
mp3 |
mp3 |
82.8 |
77.7 |
| daisy1 |
mp3 |
mp3 |
mp3 |
mp3 |
85.5 |
72.1 |
| daisy2 |
mp3 |
mp3 |
mp3 |
mp3 |
85.3 |
74.6 |
| daisy3 |
mp3 |
mp3 |
mp3 |
mp3 |
84.7 |
84.7 |
| daisy4 |
mp3 |
mp3 |
mp3 |
mp3 |
90.5 |
90.5 |
| pop1 |
mp3 |
mp3 |
mp3 |
mp3 |
74.2 |
62.7 |
| pop2 |
mp3 |
mp3 |
mp3 |
mp3 |
79.8 |
63.8 |
| pop3 |
mp3 |
mp3 |
mp3 |
mp3 |
76.8 |
62.1 |
| pop4 |
mp3 |
mp3 |
mp3 |
mp3 |
79.4 |
64.8 |
| jazz1 |
mp3 |
mp3 |
mp3 |
mp3 |
73.4 |
71.1 |
| jazz2 |
mp3 |
mp3 |
mp3 |
mp3 |
71.2 |
67.3 |
| jazz3 |
mp3 |
mp3 |
mp3 |
mp3 |
78.8 |
52.3 |
| jazz4 |
mp3 |
mp3 |
mp3 |
mp3 |
71.4 |
57.0 |
| midi1 |
mp3 |
mp3 |
mp3 |
mp3 |
74.6 |
70.7 |
| midi2 |
mp3 |
mp3 |
mp3 |
mp3 |
81.9 |
81.9 |
| midi3 |
mp3 |
mp3 |
mp3 |
mp3 |
60.0 |
60.5 |
| midi4 (Fmax=1200) |
mp3 |
mp3 |
mp3 |
mp3 |
84.8 |
73.5 |
Some results on a database I got from http://www.ee.columbia.edu/~graham/mirex_melody/, which seems to be the files "competitors" for the MIREX 2005 Melody Extraction Task could use to tune their algorithms.
The parameters are almost the same as before, except that the hopsize for the analysis windows is equal to 10ms (441 samples) to fit the given groundtruth.
| Title |
Original |
Sep.
Singer |
Sep. Music |
Remix |
PitchMatch (%) |
TotalMatch (%) |
| train01 |
mp3 |
mp3 |
mp3 |
mp3 |
80.9 | 55.1 |
| train02 |
mp3 |
mp3 |
mp3 |
mp3 |
56.9 | 37.6 |
| train03 |
mp3 |
mp3 |
mp3 |
mp3 |
77.7 | 48.3 |
| train04 |
mp3 |
mp3 |
mp3 |
mp3 |
71.2 | 61.5 |
| train05 |
mp3 |
mp3 |
mp3 |
mp3 |
76.5 | 60.0 |
| train06 |
mp3 |
mp3 |
mp3 |
mp3 |
58.7 | 29.3 |
| train07 |
mp3 |
mp3 |
mp3 |
mp3 |
72.2 | 56.4 |
| train08 |
mp3 |
mp3 |
mp3 |
mp3 |
78.7 | 58.9 |
| train09 |
mp3 |
mp3 |
mp3 |
mp3 |
85.1 |
70.4 |
... and below another example showing the possible use of our model for other instruments, in excerpts of take five (being a saxophonist, this example was compulsory for me!):
| Title |
Original |
Sep.
Saxophone |
Sep.
Background Music |
Remix |
| takefive01 |
mp3 |
mp3 |
mp3 |
mp3 |
| takefive02 |
mp3 |
mp3 |
mp3 |
mp3 |
