Thread Subject: maximum/minimum detektion in signal

i try to find maxima and minima in my data doing this:

% index to possible maxima and minima
index=2:numel(data)-1;
% maximum ist greater than its neighbours
max_index=1+find(data(index)>data(index-1) & data(index)>data(index+1))
% flip signal
data=-data;
% minima are "inverse maxima"
min_index=1+find(data(index)>data(index-1) & data(index)>data(index+1))
% flip signal again
data=-data;

this works fine (i thought so for years). I suppose that max_index and
min_index grow alternating (we have a max, then a min, then again a max
and so on).

suposing to get eg:
max_index: [4 50 103 124 156...]
min_index: [27 84 112 142 162...]

With some new data it happens that i find consecutive maxima with no
minimum between them. Any idea how this can happen? I already tried with
some kind of "tolerance" eg data(index)>data(index-1)+tol
but this did not help...
signal is sampled in 24-bit, saved as ascii and loaded from that file.

getting eg:
max_index: [4 50 103 124 156...]
min_index: [27 84 142 162...]

any hint how this can happen? Some kind of accuracy effect?
you can also help me, giving me a trick how to eliminate consecutive maxima.

tia
Lothar

Lothar Schmidt schrieb:
> i try to find maxima and minima in my data doing this:
>
> % index to possible maxima and minima
> index=2:numel(data)-1;
> % maximum ist greater than its neighbours
> max_index=1+find(data(index)>data(index-1) & data(index)>data(index+1))
> % flip signal
> data=-data;
> % minima are "inverse maxima"
> min_index=1+find(data(index)>data(index-1) & data(index)>data(index+1))
> % flip signal again
> data=-data;
>
> this works fine (i thought so for years). I suppose that max_index and
> min_index grow alternating (we have a max, then a min, then again a max
> and so on).
>
> suposing to get eg:
> max_index: [4 50 103 124 156...]
> min_index: [27 84 112 142 162...]
>
> With some new data it happens that i find consecutive maxima with no
> minimum between them. Any idea how this can happen? I already tried with
> some kind of "tolerance" eg data(index)>data(index-1)+tol
> but this did not help...
> signal is sampled in 24-bit, saved as ascii and loaded from that file.
>
> getting eg:
> max_index: [4 50 103 124 156...]
> min_index: [27 84 142 162...]
>
> any hint how this can happen? Some kind of accuracy effect?
> you can also help me, giving me a trick how to eliminate consecutive
> maxima.
>
> tia
> Lothar

just found how this can happen:

eg:
0 1 0 0 1 0 means that a max ist detected at index 2 an at index 5. No
minimum ist detected between the maxima. So far so bad...

How can i eliminate the index 124 in the example?
  max_index: [4 50 103 124 156...]
  min_index: [27 84 142 162...]

Lothar Schmidt schrieb:
> Lothar Schmidt schrieb:
>> i try to find maxima and minima in my data doing this:
>>
>> % index to possible maxima and minima
>> index=2:numel(data)-1;
>> % maximum ist greater than its neighbours
>> max_index=1+find(data(index)>data(index-1) & data(index)>data(index+1))
>> % flip signal
>> data=-data;
>> % minima are "inverse maxima"
>> min_index=1+find(data(index)>data(index-1) & data(index)>data(index+1))
>> % flip signal again
>> data=-data;
>>
>> this works fine (i thought so for years). I suppose that max_index and
>> min_index grow alternating (we have a max, then a min, then again a
>> max and so on).
>>
>> suposing to get eg:
>> max_index: [4 50 103 124 156...]
>> min_index: [27 84 112 142 162...]
>>
>> With some new data it happens that i find consecutive maxima with no
>> minimum between them. Any idea how this can happen? I already tried
>> with some kind of "tolerance" eg data(index)>data(index-1)+tol
>> but this did not help...
>> signal is sampled in 24-bit, saved as ascii and loaded from that file.
>>
>> getting eg:
>> max_index: [4 50 103 124 156...]
>> min_index: [27 84 142 162...]
>>
>> any hint how this can happen? Some kind of accuracy effect?
>> you can also help me, giving me a trick how to eliminate consecutive
>> maxima.
>>
>> tia
>> Lothar
>
> just found how this can happen:
>
> eg:
> 0 1 0 0 1 0 means that a max ist detected at index 2 an at index 5. No
> minimum ist detected between the maxima. So far so bad...
>
> How can i eliminate the index 124 in the example?
> max_index: [4 50 103 124 156...]
> min_index: [27 84 142 162...]

fixed but not yet vectorized...:
first i've made shure that the first and the last "event" is a minimum.
then:

     for i=1:numel(max_index)
         while i<numel(max_index) && max_index(i)<min_index(i)
             max_index(i)=[];
         end
         while i<numel(min_index) && min_in dex(i+1)<max_index(i)
             min_index(i+1)=[];
         end
     end

any idea how to vectorize?

Lothar Schmidt:
<SNIP min-max-only, please

one of the solutions

% the data
     a=[4,50,103,124,156];
     b=[27 84 142 162];
% the engine
     [as,ix]=sort([a,b]);
     ix=[true,diff(ix)~=1];
     as=as(ix);
     an=as(1:2:end);
     bn=as(2:2:end);
% the result
     disp(an);
     disp(bn);

us

us schrieb:
> Lothar Schmidt:
> <SNIP min-max-only, please
>
> one of the solutions
>
> % the data
> a=[4,50,103,124,156];
> b=[27 84 142 162];
> % the engine
> [as,ix]=sort([a,b]);
> ix=[true,diff(ix)~=1];
> as=as(ix);
> an=as(1:2:end);
> bn=as(2:2:end);
> % the result
> disp(an);
> disp(bn);
>
> us

your my hero! I tried something similar but got stuck in index
confusion... thanks a lot

Lo