Am 3/11/2014 6:51 PM, schrieb udara de silva: > Greetings all, > I have an project idea for Mixxx which I think is* very very important *for > DJs. That is MMD Algorithms for make track selection super easy. First let > me explain what i mean by mood. Well I'm feeling exciting now. That's a > mood. I feel like dancing,rocking,clubbing or I'm feeling romantic,sad or > relaxing. These are what I mean by mood. How is this quantified? BPM, spectrum, timbre? There has been a lot of research on that part that you should read before thinking further. The thesis to the mirage plugin for banshee is a good start
http://hop.at/mirage/mirage.pdf Also we had project like that already last year and I plan to include it in the 1.13 version http://mixxxblog.blogspot.de/2013/05/google-summer-of-code-2013-projects.html > Say your are DJying a dance floor > so you want to mix tracks that make people on the dance floor feel like > dancing. Well you would say a good DJ should know that track. True enough > but having more than 10,000 songs collection what would be the time taken > to select the track on the line. *My proposal is to create an Artificial > Intelligent to detect mood of the music* and generate a playlist out of > those 10000 songs and then the DJ can easily figure out what track does he > need to play. Let me discuss on how to implement the system in brief, > > before starting off let me post a link again for an artificial intelligence > I created who I call "megamind" [?] http://www.pixelzexplorer.org/megamind/ > you > can ask a question and hit enter to see his answer. There I use natural > language processing to determine what the user might be wanting to know > > Having said that all let me move on to my implementation. With max replies > I broken my proposal to 4 phases and they goes as follows. > > *Phase 1 : **Supervised mood detection* > This is quite easy. the plan is to add UI elements to let the DJ choose the > proper mood of the song. This would be then saved with the song so that > later this information will be used to create the play list. This would be > the start point for the project but practically setting moods of all tracks > would be quite difficult since there will be lot of songs. What are your features / target function for the ML algorithm? What kind of model do you want to use? > > *Phase 2 : **Unsupervised learning for mood detection* > This would be the most cool part and hence is the most difficult. In this > track mood will be automatically detected. Is this possible??? YES. I would > love to go into details how this would be done if you guys are interested > in accepting my proposal. But what should be noted is that there is no > perfect algorithm for this because the concept of mood is ambiguous. What > we could build is an algorithm based of Fuzzy Logic. But I am very > convinced this is something possible to do. The same question as above. What would you use as an input? This is a seriously though problem. Thinks that work for stuff like house usually don't work for other genres. If you want us to take you seriously here you should show that you have read the literature. Although I'm not to optimistic in taking this one because we already have such a feature lined up for the future > > *Phase 3 : **Combination of the supervised and unsupervised information* > This would also be an easy task. If the DJ think automatically detected > mood is not correct in this phase there will be a feature to change it in > the desirable way. > > *Phase 4 : **Automatically generating the playlist* > >From what Max told me there has being some work done in this area so > implementing this would be quite easy. Since I trust Max [?] I would rank > this phase as an easy one. Otherwise I would have to work on this one too.. > > So this is my proposal. I have done some work in NLP for decision making > but haven't try to use AI on the field of Music. But I love challenges and > would love to tackle the problem. I have work with Java,C++,C and has a > background knowledge on QT as well. This is my first attempt to contribute > my skills to the Opensource community and wish you guys let me in. Thank > you...! > > > > ------------------------------------------------------------------------------ > Learn Graph Databases - Download FREE O'Reilly Book > "Graph Databases" is the definitive new guide to graph databases and their > applications. Written by three acclaimed leaders in the field, > this first edition is now available. Download your free book today! > http://p.sf.net/sfu/13534_NeoTech > > > _______________________________________________ > Get Mixxx, the #1 Free MP3 DJ Mixing software Today > http://mixxx.org > > > Mixxx-devel mailing list > Mixxx-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/mixxx-devel ------------------------------------------------------------------------------ Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/13534_NeoTech _______________________________________________ Get Mixxx, the #1 Free MP3 DJ Mixing software Today http://mixxx.org Mixxx-devel mailing list Mixxx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mixxx-devel
