noise removal from audio signal using cepstral analysis

hello,
i have an audio signal mixed with white noise. How do i use the cepstrum
technique to remove the noise from it.

please tell me from where should i start. i want to know the basics of
cepstrum and how can i use it to fullfill my purpose.

please guide me about the literature which i should read.

thanks

On Jan 21, 2:48=A0am, "enricophpdsp"
<enricophpdsp@n_o_s_p_a_m.yahoo.com> wrote:
> hello,
> i have an audio signal mixed with white noise. How do i use the cepstrum
> technique to remove the noise from it.
>
> please tell me from where should i start. i want to know the basics of
> cepstrum and how can i use it to fullfill my purpose.
>
> please guide me about the literature which i should read.
>
> thanks

Why do you want to apply cepstral methods to this problem?  What sorts
of signals does the cepstrum decompose?

Dirk

enricophpdsp schrieb:
> hello,
> i have an audio signal mixed with white noise. How do i use the cepstrum
> technique to remove the noise from it.
>
> please tell me from where should i start. i want to know the basics of
> cepstrum and how can i use it to fullfill my purpose.
>
> please guide me about the literature which i should read.
>
> thanks

The cepstrum is used for analysis and classification of speech but it 
does not help you to remove noise. If your signal exhibits a low SNR you 
need to do noise removal before computing a cepstrum. As far as I can 
remember the Cepstrum expects a 'normalized' input anyways. Google for 
Spectral Subtraction or Wiener Filters.

Greetz,

Sebastian

On Jan 22, 5:37=A0am, Sebastian Doht <seb_d...@lycos.com> wrote:
> enricophpdsp schrieb:
>
> > hello,
> > i have an audio signal mixed with white noise. How do i use the cepstru=
m
> > technique to remove the noise from it.
>
> > please tell me from where should i start. i want to know the basics of
> > cepstrum and how can i use it to fullfill my purpose.
>
> > please guide me about the literature which i should read.
>
> > thanks
>
> The cepstrum is used for analysis and classification of speech but it
> does not help you to remove noise. If your signal exhibits a low SNR you
> need to do noise removal before computing a cepstrum. As far as I can
> remember the Cepstrum expects a 'normalized' input anyways. Google for
> Spectral Subtraction or Wiener Filters.
>
> Greetz,
>
> Sebastian

It can actually.

http://ieeexplore.ieee.org/Xplore/login.jsp?url=3Dhttp%3A%2F%2Fieeexplore.i=
eee.org%2Fiel5%2F5465897%2F5472440%2F05472686.pdf%3Farnumber%3D5472686&auth=
Decision=3D-203

Hardy

HardySpicer schrieb:
> On Jan 22, 5:37 am, Sebastian Doht<seb_d...@lycos.com>  wrote:
>> enricophpdsp schrieb:
>>
>>> hello,
>>> i have an audio signal mixed with white noise. How do i use the cepstrum
>>> technique to remove the noise from it.
>>
>>> please tell me from where should i start. i want to know the basics of
>>> cepstrum and how can i use it to fullfill my purpose.
>>
>>> please guide me about the literature which i should read.
>>
>>> thanks
>>
>> The cepstrum is used for analysis and classification of speech but it
>> does not help you to remove noise. If your signal exhibits a low SNR you
>> need to do noise removal before computing a cepstrum. As far as I can
>> remember the Cepstrum expects a 'normalized' input anyways. Google for
>> Spectral Subtraction or Wiener Filters.
>>
>> Greetz,
>>
>> Sebastian
>
> It can actually.
>
> http://ieeexplore.ieee.org/Xplore/login.jsp?url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F5465897%2F5472440%2F05472686.pdf%3Farnumber%3D5472686&authDecision=-203
>
> Hardy

Thanks for the link. Well, I do not have access to this paper and so I 
am just trying to deduct from the abstract, that they probably found a 
way to do operations in the cepstral domain which are usally applied in 
the correlation domain. Sounds to me like making things more complicated 
than required but if it saves you some dB it may be worth the effort.

>hello,
>i have an audio signal mixed with white noise. How do i use the cepstrum
>technique to remove the noise from it.
>
>please tell me from where should i start. i want to know the basics of
>cepstrum and how can i use it to fullfill my purpose.
>
>please guide me about the literature which i should read.
>
>thanks
>


hello,
actually i m participating in an even related to digital signal processing.
in that even , the problem statement is not declared, but they have
indicated that the participants have to use cepstrum analysis . 
so please someone tell me that what literature should i read and how should
i proceed. thanks for the valuable replies.

On Jan 23, 12:54=A0am, Sebastian Doht <seb_d...@lycos.com> wrote:
> HardySpicer schrieb:
>
>
>
> > On Jan 22, 5:37 am, Sebastian Doht<seb_d...@lycos.com> =A0wrote:
> >> enricophpdsp schrieb:
>
> >>> hello,
> >>> i have an audio signal mixed with white noise. How do i use the cepst=
rum
> >>> technique to remove the noise from it.
>
> >>> please tell me from where should i start. i want to know the basics o=
f
> >>> cepstrum and how can i use it to fullfill my purpose.
>
> >>> please guide me about the literature which i should read.
>
> >>> thanks
>
> >> The cepstrum is used for analysis and classification of speech but it
> >> does not help you to remove noise. If your signal exhibits a low SNR y=
ou
> >> need to do noise removal before computing a cepstrum. As far as I can
> >> remember the Cepstrum expects a 'normalized' input anyways. Google for
> >> Spectral Subtraction or Wiener Filters.
>
> >> Greetz,
>
> >> Sebastian
>
> > It can actually.
>
> >http://ieeexplore.ieee.org/Xplore/login.jsp?url=3Dhttp%3A%2F%2Fieeexplo.=
..
>
> > Hardy
>
> Thanks for the link. Well, I do not have access to this paper and so I
> am just trying to deduct from the abstract, that they probably found a
> way to do operations in the cepstral domain which are usally applied in
> the correlation domain. Sounds to me like making things more complicated
> than required but if it saves you some dB it may be worth the effort.

There is a whole literature on noise reduction in the cepstral domain.
Much of it revolves around AR modelling, but not all of it.

Hardy