# "Extracting" sound between a certain frequency-range / How to interpret FFT Output

This topic is 1134 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

Hey guys!

For my newest project I need to be able to identify different frequencies in a specific sound-source (Either the direct input, or some sound file like mp3 or something). I don't know a single thing about sound processing and signal processing and all that stuff, but I hope that I will be able to explain myself.

Here's what I'd like to achieve:

• From a "complete" sound-source such as a mp3-file, I need to be able to analyze it's frequencies in order to understand in which frequency-range a specific sound I recorded lies.
• From some direct-input OR a complete sound-source such as a mp3-file, I need to be able to:
• - Extract sound of a frequency between my previously analyzed frequency-range in order to verify my previous analysis.
• - Detect sound of a frequency between my previously analyzes frequency-range from some continuous input, and somehow do something when it got detected.

As I said, I don't know anything about signal processing, but I heard that a Fast Fourier Transform outputs data that can be interpreted to somehow separate the different frequencies in a sound-file. With Mathematica I tried to apply this algorithm to one of my files, and got a result - But I am not able to interpret it.

http://i.stack.imgur.com/9Meca.jpg

I know that the output should always be symmetric if real input was given, and that, in my case, the frequencies go from 0Hz to 22k-and-something Hz, but else I don't know anything. I think that I am pretty confused by the fact that there isnt any "time-basis": So I don't know how to know what frequency was present at which point of the sound. Ultimately, I'd like to represent the frequency(ies) in a specific range by some sinusoid-like-curve (I know that the frequency that I should get has a peculiar motive, but I dont know how to get it. The sound is the one of insects by the way.)

Help would be greatly appreciated.

Kind regards

##### Share on other sites
You are right that the time information is lost when you do the FFT. If you want to know when a particular frequency was active, I can think of two methods:
(1) Divide the signal into time windows (it's OK if they overlap) and compute the FFT of each window. See here.
(2) Use the continuous wavelet transform or some other time-frequency representation of the signal.

Although (2) is probably a better mathematical tool, you said you know nothing about signal processing, so I suggest you stick to (1).

EDIT: You should probably also take a look at the Wikipedia page on spectrograms. Edited by Álvaro

1. 1
Rutin
46
2. 2
3. 3
4. 4
5. 5
JoeJ
19

• 13
• 10
• 12
• 10
• 13
• ### Forum Statistics

• Total Topics
632998
• Total Posts
3009808
• ### Who's Online (See full list)

There are no registered users currently online

×