how to divide sample rate frequency by 10 bands for equalization

Question

How to divide sample rate frequency by 10 bands for equalization. For example i have audio pcm with 44100 hz and 1 byte per sample. How can i understand which sample is related with specified band. And how to divide samples to bands.

Can you give me explanation or link.

Answer 1

There are two common approaches for this:

1. Use a filter bank. You need to design 10 band-pass filters. All the filters receive the full bandwudth input and the output of each filter is a sub-band containing only the frequencies of interest for that band. For equalization purposes you can then scale the amplitude of each filter output and then re-combine the scaled outputs.

2. Use FFT + IFFT. Use a (forward) FFT to convert your time domain data into the frequency domain. Modify the amplitudes in each frequency domain bin. Apply inverse FFT to get equalized data back into the time domain. You'll need to choose a sufficiently large FFT size (N) to get the required frequency resolution for your equalization bands. You'll also need to use an overlap technique to avoid discontinuities between successive frames.

Answer 2

Do a search on "quadrature mirror filter" or "perfect reconstruction". This paper looks to be as good a place as any to begin:

http://authors.library.caltech.edu/5463/1/VAIieeeasspm87.pdf