My SDP! Music Classification ^^

How many times have you thought of actually being able to classify the song you are listening to into a specific genre? Countless. It is probably like climbing the Everest for even the most accomplished musicians out there. But the Gurus of music have actually found ways to make classes in their musical types. One of which is the most prominent is the classification of guitar/drum music according to the "scales" that they play. If you’re a musician you would know that a scale is actually a progression of a definite sequence of "notes". I play a few notes on my 6-string, say ’A B B B (octave) E G’, will form a scale. Using the classification according to scales by the intenational music community, they have actually seperated thousands of scales that make up genre’s like Country, Rock, Metal, Soft Rock, Hard Rock, Rock and Roll, etc.

What my project does, is that it takes input from your guitar, and breaks it down into the constituent notes. This is called ’segmentation’. Then after this these ’segmented’ notes are filtered for noise, and various other multiples of the signal input. These filtered notes are then passed on to a Neural Network, that works pretty much like a brain and trys to recognize the note that it has been told to recognize. It does so with 85% to 90% acuracy. Which is kinda good for the kind of conditions that we were conducting the experiments and tests in. Full of (background) noise and reverberation. The frequencies and amplitudes got all messed up. Note recognition helps in Scale recognition. Scale recognition is where we determine the genre of the music being played. Below you will see a picture containing the note signal, then its Fourier Transform, and then the Spectral view of the Fourier Transform (which elaborates the power contribution to the signal of a certian frequency).

Note E
I am currently working on the documentation of this project. Some stage/progress reports are already made, but I have no intention of uploading them so soon. Only after the Final Project Report is done. InshAllah that will be done soon. The report will also include respective acknowledgements and references.

Come to think of the applications of this system, and they are emmense. Only looking at the needs of modren society and music, we see that music piracy is one of the most rampant crimes. File swapping is one of the main concerns of the Music and record labels out there. The contemporary ways of dealing with that are file name and extension security or water marking, both of wich are easily removed by and intelligent user. What they cant change is the music in the file itself. Which is where we gain strength from and our project uses the musical characteristic of that piece to keep it secure. Secondly, most of the time we are humming a tune to a song to which we do not know the name of, or the artist. If we are able to play it to our system, we will be able to identify the artist and song name using the feature extraction methods we have employed. Apart from the music industry there are a zillion more uses to Sound Pattern Recognition Methods in fields where waves are used. Radiology, Astronomy, Cosmology, Telecommunications and etc.

So keep thinkin, post your comments if you like my porject. And enhancment suggestions are welcome. If you think that you have already worked on this and your work has be used in someway, dont worry, I am compiling a acknowledgement list and it will have your name.