The detecting process of FFT matrix approach 1.1 2006/12/23 07:28:41.888 US/Central 2007/09/25 10:12:35.067 GMT-5 Yiming Wang yw4630@rice.edu Yiming Wang yw4630@rice.edu Rice University ELEC 301 md@rice.edu Describe the detailed information about the detecing process of the FFT matrix approach.

The Detecting Process: So we load the song we want to detect into Matlab. The use would just type Genre (‘song’). After that, our program would do the same process of windowing; it did to the songs in the database (FFT of windows of 15ms length). This will leave us with another, smaller matrix that represents the song we want to test. We basically perform an operation of dot production for each column of the tested song with our data base. ( The operation of Dot Product gives a bigger value for a vector that is similar, and a small value for one that is different ). So we basically check for the max. value of each dot product operation. And this means that this column (frequencies) of the tested song is really close to this column that represent a certain genre. We then ask Matlab about the genre this column represented. And save this value. After doing this for all the columns of the tested song, we’ll end up with a bunch of numbers going from (1-5). Finally we check which number is the most frequent in the group, and this will tell us what genre the tested song is most likely to be.