PLEASE READ EXPLANATION
AND NOTES BEFORE LAUNCHING APPLETS
applet 1 - applet 2 - applet 3
This applet is a genre guessing game.
Aims:
-This applet is similar to the experiment done by Perrot and Gjerdigen (1999) which suggests that humans can judge a genre by using only immediate attributes (a combination of timbre, texture, rhythm, instrumentation...?) and without constructing higher level features (they don't have time to do that).
-This applet reflects the idea that recognition in the prescriptive approach is based on small frames, (250ms to 1 sec), and raises an interesting analogy with the experiment done by Perrot and Gjerdigen (1999).
-This applet introduces samples of different genres, the latter which can be used in applet 3. Applet 3 extracts features (timbre information based on FFT coefficients..), hence the user can observe how the features vary with the genres.
-Using this applet, the user might notice how difficult it is to distinguish Pop from Rock (as noted in the study by the authors of [2])
-This applet gives an idea of manual classification (discussed at the beginning of the tutorial), often used in conjunction with automatic recognition.
-Finally, with this applet, you also have fun, and can practice your music genre recognition talents!
Instructions:
-Click on PlayAudio button to start.
-Hear the sample and select a choice from the list.
-Click on Validate answer. Repeat procedure for next 10 titles.
-You have the choice to repeat the same game with the same set of 10 samples or play a new game with new samples.
NOTE1: on some computers, clicking on the play button several times without waiting for 5 to 10 seconds between each click might not play the clip.
NOTE2: If you wonder how the manually chosen 1 second samples were assigned to a genre, well I assigned the genre based on first the majority of the music websites on the internet, second, on my musical and music history knowledge. Again, the whole tutorial carried the idea that genre is controversial and highly subjective...
By: Imad Khoury
This applet is a subgenre guessing game.
Aims:
-This has the same aims as applet 1 with one additional aim:
-It shows the difficulty of recognizing subgenres as compared to genres, which is exactly reflected in automatic genre recognition systems, as discussed in the tutorial.
Instructions:
-Click on PlayAudio button to start.
-Hear the sample and select a choice from the list.
-Click on Validate answer.
-You have the choice to repeat the same game with the same set of 10 samples or play a new game with new samples.
NOTE: on some computers, clicking on the play button subsequent times without waiting for 5 to 10 seconds between each click might not play the clip.
By: Imad Khoury
This applet is a music feature extraction applet, as a first step towards genre recognition:
Aims:
-This applet offers to the user some possibilities of extracting features (FFT coefficients) from audio clips, a step used in the prescriptive approach as we have seen.
-One could compare the FFTs of different genres by selecting songs of different genres..
-Also, one could experiment and find that many songs of different genres can have the same FFTs and vice versa, as previously noted in the tutorial
Instructions:
-Try to use simple audio waveforms using oscillator and explore the spectrum FFT coefficients (plotted)
-You can add noise or a DC (constant value)
-Then make sure to click on 'use URL' :
-Once you click on that, you can either enter a url to a .wav or select one from the list above (which contains all the 1 sec samples u saw in game 1 and 2).
-Press 'Open'
-Then press 'Plot'. Observe the extracted FFT coefficients. Observe for example that if you choose the 'FUGUE 10 BY BACH' the FFT coefficients are nearly zero everywhere except for a given frequency where you have a peak (for the entire 1 second frame), whereas if you choose an 'AC DC' song the FFT coefficients are high on a bigger range of frequencies, and most are concentrated in the bass range
-Click plot several times to move to the right in the samples axis, or use slider provided.
NOTE: You can select any sample from the samples you heard in games 1 and 2. Also, you can use your own audio clips. HOWEVER, since an applet cannot access local resources, you will need to download the .class files, put them in a folder, and then put your own audio clips in that same folder so that the applet will be able to load them.
By: Tsan-Kuang Lee. Modified by Imad Khoury. free software.