1
$\begingroup$

I am new in signal processing and I want to extract the talking (lyrics) of a person from a sound so I can analyze it ;another application would be if that person is talking and there are many sounds behind him the applications will detect his sound only.

I have read many articles about single processing like (analog to numeric,FFT,Gaussian, FTD , Correlation ..... ) but I can't combine those all information together because i am new in this domain like i said before.

please someone help me I am totally lost ---> sorry for my English <----

$\endgroup$

closed as too broad by A_A, hotpaw2, MBaz, lennon310, Tendero Mar 14 '18 at 16:55

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ Did you see the article I sent you? $\endgroup$ – VMMF Mar 9 '18 at 16:32
0
$\begingroup$

What you are trying to do might be extremely difficult depending on numerous factors. I have read about a technology called blind source de-convolution which tries to separate one audio source from others. In the end rather than recovering every audio source you will most likely recover only the one you are interested in.

I would suggest you start with this survey article which lists many of the attempts that have been done to achieve blind source de-convolution.

If what your are trying to do doesn't have to be too precise, as a very simple start you might just want to filter every frequency component above 4-5 kHz because the main energy of human voice is below that. However if it is a song with a strong bass, for instance hip-hop, pop, drums, etc, this won't help you clean that

$\endgroup$
  • $\begingroup$ Great ,yeah I have read it and that will totally help me , but I am still confused little bit about it, what I understand for now is : the sound will be translated from analog to numeric by the sound sensor , So the computer will save it as (01010101010) data , to visualize that data as a graph we use the FFT algorithme am I right in that ?? $\endgroup$ – Badis vadici Mar 11 '18 at 9:08
  • $\begingroup$ Actually, the sound will be translated from analog to digital by the digital to analog converter placed after the sound sensor. Yes the computer will save it in bits, typically 16 bits per sample. So you could interpret that as short. The FFT will allow you to visualize the frequency components that are present in the sound during the time you recorded. If you want to have an idea of what frequencies are present at which time use an spectrogram (en.wikipedia.org/wiki/Spectrogram ). However, knowing all this doesn't in principle mean that you can separate the audio sources $\endgroup$ – VMMF Mar 12 '18 at 14:07

Not the answer you're looking for? Browse other questions tagged or ask your own question.