d91
Newbie level 1
Hi people..
I am thinking about noise reduction in skype.. Like.. the standard annoying buzz the microphone usually gives, or something that is always there, but you can't get rid off.
I am using matlab and made 2 samples, 1 sample containing the noise + speech, and 1 sample containing only the noise.
I made a spectrogram off that and.. the frequency of the noise (and the intensity) is very much in the speech bandfrequency, which makes it ofcourse very hard to filter..
The idea I have is (all static now, but for an application later) take a sample of a person talking for about 3 seconds. Then a sample of the person not talking for also about 3 seconds.
Then, subtract the current stream of voice (I hope you get what i mean) with the sample of noise that is recorded..
Is this a good idea to try?
Or does anyone have an idea to get the noise out of a sample, without distorting the voice.. (the literature I found couldn't help me too much on it :/)
Another idea is to make a bandpass filter, which passes frequencies from 300 Hz to about 3 KHz, but ofcourse if the noise is within the 300 Hz - 3KHz range, then the problem still remains..
I am not an expert on DSP but I have some basic knowledge I guess, hopefully enough to fix this =)
Also, this is for a school assignment and a classmate and myself came up with this project.
Thanks in advance.. hopefully I have a reaction soon
Davey
I am thinking about noise reduction in skype.. Like.. the standard annoying buzz the microphone usually gives, or something that is always there, but you can't get rid off.
I am using matlab and made 2 samples, 1 sample containing the noise + speech, and 1 sample containing only the noise.
I made a spectrogram off that and.. the frequency of the noise (and the intensity) is very much in the speech bandfrequency, which makes it ofcourse very hard to filter..
The idea I have is (all static now, but for an application later) take a sample of a person talking for about 3 seconds. Then a sample of the person not talking for also about 3 seconds.
Then, subtract the current stream of voice (I hope you get what i mean) with the sample of noise that is recorded..
Is this a good idea to try?
Or does anyone have an idea to get the noise out of a sample, without distorting the voice.. (the literature I found couldn't help me too much on it :/)
Another idea is to make a bandpass filter, which passes frequencies from 300 Hz to about 3 KHz, but ofcourse if the noise is within the 300 Hz - 3KHz range, then the problem still remains..
I am not an expert on DSP but I have some basic knowledge I guess, hopefully enough to fix this =)
Also, this is for a school assignment and a classmate and myself came up with this project.
Thanks in advance.. hopefully I have a reaction soon
Davey