FPGA Central

Hi,

Designing filters that remove frequency components from the entire
signal is easy.

But how do you break a signal in 2 parts and do some frequency-based
filtering on 1 part of it only?

I need to design a filter that to removes noise from beats 4-8 of an
8-beat ECG signal. Need to leave beats 1-3 alone as they are clean.

Thanks,

Parag

Parag wrote:
> Hi,
>
> Designing filters that remove frequency components from the entire
> signal is easy.
>
> But how do you break a signal in 2 parts and do some frequency-based
> filtering on 1 part of it only?
>
> I need to design a filter that to removes noise from beats 4-8 of an
> 8-beat ECG signal. Need to leave beats 1-3 alone as they are clean.
>
> Thanks,
>
> Parag

Use two signal paths, one filtered, the other not. Ensure that the paths
have the same delay. Take your output from the filtered or the
unfiltered path as appropriate.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Jerry Avins wrote:
> Parag wrote:
> > Hi,
> >
> > Designing filters that remove frequency components from the entire
> > signal is easy.
> >
> > But how do you break a signal in 2 parts and do some
frequency-based
> > filtering on 1 part of it only?
> >
> > I need to design a filter that to removes noise from beats 4-8 of
an
> > 8-beat ECG signal. Need to leave beats 1-3 alone as they are clean.
> >
> > Thanks,
> >
> > Parag
>
> Use two signal paths, one filtered, the other not. Ensure that the
paths
> have the same delay. Take your output from the filtered or the
> unfiltered path as appropriate.

Agreed. Remember to use a linear-phase filter if the time interleaving
is to work correctly (otherwise the "delay" of the filtered signal is
not defined). You will have disontinuities when you hard-switch between
the filtered and the clean signal --- try crossfading if this is a
problem.

Regards,
Andor

I am no authority on biological signals. But assuming that the heart
beats occure at precise regularity, you can split the data into two
streams. The difficult question is: how do you measure time period
correctly. I dont know the answer, but can think of something if I get
a look at the data/plot. Idealy, I would like to use some simple
mechanism, like a correlator to find out where the next period of the
heart-beat is starting. But thats just a wild guess


"Parag" <[email protected]> wrote in message news:<[email protected] roups.com>...
> Hi,
>
> Designing filters that remove frequency components from the entire
> signal is easy.
>
> But how do you break a signal in 2 parts and do some frequency-based
> filtering on 1 part of it only?
>
> I need to design a filter that to removes noise from beats 4-8 of an
> 8-beat ECG signal. Need to leave beats 1-3 alone as they are clean.
>
> Thanks,
>
> Parag