Useful audio data from FFT

Question

I'm trying to do a simple music visualization in java. I have two threads set up, one for playing the clip, and another for extracting a chunk of bytes from the clip to process with an FFT. The processed array can then be sent to the JFrame that will handle drawing, and used as a parameter for some sort of visual.

I'm not exactly sure what to do with the data, however. I've been just using a power spectrum for now, which gives me very limited response, and I realize is too general for what I am trying to do. I'm open to using any FFT library out there, if there is a specific one that will be especially helpful. But, in general, what can I get from my data after doing an FFT, and how can I use to show decently accurate results in the visuals?

hotpaw2 · Accepted Answer · 2011-02-20 06:14:52Z

All FFTs will do pretty much the same thing given the same data. The FFT parameters you can vary are the scale factor, the length of the FFT (longer will give you higher frequency resolution, shorter will give you better time response), and (pre)windowing the data, which will cause less "splatter" or spectral leakage of spectral peaks. You can zero-pad an FFT for interpolating smoother looking results. You can average the magnitude results of several successive FFTs to reduce the noise floor. You can also use a scaling function such a log scaling (or log log, e.g. log on both axis) for presenting the FFT magnitude results.

The phase of a complex FFT is usually unimportant for any visualization unless you are doing some type of phase vocoder analysis+resynthesis.

Collectives™ on Stack Overflow

Useful audio data from FFT

1 Answer 1

Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Related