Forums

speech extraction

Started by vishnuj17 May 8, 2011
Hi.. I was trying to extract speech data from raw data with background
traffic. i tried wiener filter, but was not so good.
Could you share with me your suggestions that will help me to extract
speech only from background noise

Thank you


On 05/08/2011 05:12 PM, vishnuj17 wrote:
> Hi.. I was trying to extract speech data from raw data with background > traffic. i tried wiener filter, but was not so good. > Could you share with me your suggestions that will help me to extract > speech only from background noise
If by "speech" you mean the voice signal, then you might want to search for "VAD" ("voice activity detection") and "spectral subtraction." -- Randy Yates % "So now it's getting late, Digital Signal Labs % and those who hesitate mailto://yates@ieee.org % got no one..." http://www.digitalsignallabs.com % 'Waterfall', *Face The Music*, ELO
Randy Yates wrote:
> On 05/08/2011 05:12 PM, vishnuj17 wrote: >> Hi.. I was trying to extract speech data from raw data with background >> traffic. i tried wiener filter, but was not so good. >> Could you share with me your suggestions that will help me to extract >> speech only from background noise > > If by "speech" you mean the voice signal, then you might want to > search for "VAD" ("voice activity detection") and "spectral > subtraction."
And if he was trying to extract "intelligence" ? ???
On 05/08/2011 06:26 PM, Richard Owlett wrote:
> Randy Yates wrote: >> On 05/08/2011 05:12 PM, vishnuj17 wrote: >>> Hi.. I was trying to extract speech data from raw data with background >>> traffic. i tried wiener filter, but was not so good. >>> Could you share with me your suggestions that will help me to extract >>> speech only from background noise >> >> If by "speech" you mean the voice signal, then you might want to >> search for "VAD" ("voice activity detection") and "spectral >> subtraction." > > And if he was trying to extract "intelligence" ? ???
Tell Scotty to beam him up. -- Randy Yates % "So now it's getting late, Digital Signal Labs % and those who hesitate mailto://yates@ieee.org % got no one..." http://www.digitalsignallabs.com % 'Waterfall', *Face The Music*, ELO
On May 9, 9:12&#2013266080;am, "vishnuj17" <vishnuj17@n_o_s_p_a_m.gmail.com>
wrote:
> Hi.. I was trying to extract speech data from raw data with background > traffic. i tried wiener filter, but was not so good. > Could you share with me your suggestions that will help me to extract > speech only from background noise > > Thank you
How many sensors do you have? How are they arranged? Look up ICA independent component analysis or Blind source separation. You'll normally need n sensors for n signals. Hardy
>On 05/08/2011 05:12 PM, vishnuj17 wrote: >> Hi.. I was trying to extract speech data from raw data with background >> traffic. i tried wiener filter, but was not so good. >> Could you share with me your suggestions that will help me to extract >> speech only from background noise > >If by "speech" you mean the voice signal, then you might want to >search for "VAD" ("voice activity detection") and "spectral >subtraction." >-- >Randy Yates % "So now it's getting late, >Digital Signal Labs % and those who hesitate >mailto://yates@ieee.org % got no one..." >http://www.digitalsignallabs.com % 'Waterfall', *Face The Music*, ELO >Thank you for the response..
I tried ordinary VAD which checked the avg amplitude of a BPF tuned for speech frequencies. But also didnt turned out to be much useful, since the background traffic noise is also in the same freq band.
>On May 9, 9:12=A0am, "vishnuj17" <vishnuj17@n_o_s_p_a_m.gmail.com> >wrote: >> Hi.. I was trying to extract speech data from raw data with background >> traffic. i tried wiener filter, but was not so good. >> Could you share with me your suggestions that will help me to extract >> speech only from background noise >> >> Thank you > >How many sensors do you have? How are they arranged? Look up ICA >independent component analysis >or Blind source separation. You'll normally need n sensors for n >signals. > >Hardy >
I hope i can do that from and ordinary sound recorder, just a mic must be enough.
>Randy Yates wrote: >> On 05/08/2011 05:12 PM, vishnuj17 wrote: >>> Hi.. I was trying to extract speech data from raw data with background >>> traffic. i tried wiener filter, but was not so good. >>> Could you share with me your suggestions that will help me to extract >>> speech only from background noise >> >> If by "speech" you mean the voice signal, then you might want to >> search for "VAD" ("voice activity detection") and "spectral >> subtraction." > >And if he was trying to extract "intelligence" ? ??? > > >
Yes, i also think that there must be some intelligence that will follow human speech pattern. Now i am trying if prediction algorithms will help me. but getting no idea how to implement them in this scenario
>On 05/08/2011 05:12 PM, vishnuj17 wrote: >> Hi.. I was trying to extract speech data from raw data with background >> traffic. i tried wiener filter, but was not so good. >> Could you share with me your suggestions that will help me to extract >> speech only from background noise > >If by "speech" you mean the voice signal, then you might want to >search for "VAD" ("voice activity detection") and "spectral >subtraction." >-- >Randy Yates % "So now it's getting late, >Digital Signal Labs % and those who hesitate >mailto://yates@ieee.org % got no one..." >http://www.digitalsignallabs.com % 'Waterfall', *Face The Music*, ELO >
also since the signal band and noise band are overlapped, ordinary spectral subtraction will damage the signal also.
On May 9, 10:31=A0am, "vishnuj17" <vishnuj17@n_o_s_p_a_m.gmail.com>
wrote:
> >On 05/08/2011 05:12 PM, vishnuj17 wrote: > >> Hi.. I was trying to extract speech data from raw data with background > >> traffic. i tried wiener filter, but was not so good. > >> Could you share with me your suggestions that will help me to extract > >> speech only from background noise > > >If by "speech" you mean the voice signal, then you might want to > >search for "VAD" ("voice activity detection") and "spectral > >subtraction." > >-- > >Randy Yates =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0% "So now it's ge=
tting late,
> >Digital Signal Labs =A0 =A0 =A0 =A0 =A0 =A0 =A0% =A0 =A0and those who he=
sitate
> >mailto://ya...@ieee.org =A0 =A0 =A0 =A0 =A0% =A0 =A0got no one..." > >http://www.digitalsignallabs.com% 'Waterfall', *Face The Music*, ELO > > also since the signal band and noise band are overlapped, ordinary spectr=
al
> subtraction will damage the signal also.
Not necessarily true. It depends on the SNR. Do a google search on _spectral subtraction speech enhancement_. Also, if you don't get the parameters correct, you will defeat your purpose and screw up the signal. For a _basic_ discussion of the parameters and how to set them for a basic spectral subtraction noise reduction system, look up the comp.dsp 2004 conference and download a presentation on noise reduction techniques.