easy start of interpolation question.

[Roderik Emmerink]

Hi,

I want to determine if a next found point is near to the previous found
point and so determine of I should count it or not. In my application it
often happens that their are wrong points in my signal. When I have the
first points when measuring I can determine where the next point could
be. So, my problem is, how to determine the first points (when there is
no history yet). I hope someone will have a simple answer to this (I
don't want huge algorithms, since it is already a slow application).

Kind regards,

Roderik

[Bernhard Holzmayer]

Roderik Emmerink wrote:

> Hi,
>
> I want to determine if a next found point is near to the previous
> found point and so determine of I should count it or not. In my
> application it often happens that their are wrong points in my
> signal. When I have the first points when measuring I can
> determine where the next point could be. So, my problem is, how to
> determine the first points (when there is no history yet). I hope
> someone will have a simple answer to this (I don't want huge
> algorithms, since it is already a slow application).
>
> Kind regards,
>
> Roderik

Hi Roderik,

it depends on your incoming data.
If you don't have the history, try to evaluate the future ;-)

I'd try this:
- store incoming data into a buffer
- as soon as the buffer is full enough, evaluate the data in it
backwards.
it should be possible to retrieve the first points which you're
looking for.
Then take the stored data and correct them as you did.

Disadvantage is that your incoming data are delayed until you have
the first points.
And if you have a synchronous stream, you'll not catch up, so they
will have a constant delay.

Advantage of this approach is that you need not blow up your basic
algorithm, instead you'd use the additional calculation only during
the startup phase, perhaps in an "if" clause.

Bernhard

--
before sending to the above email-address:
replace deadspam.com by foerstergroup.de