Detecting and saving the values of a single pulse signal form the continuous incoming pulses signals

Hello, I am getting data of pulse signals into matlab through serial port.I want to detect and distinguish each pulse pulse signal and want to save the data(all the samples of amplitude that can reconstruct the signal for the later use) of the pulse signals.

As all pulse signals don't have the same number of samples it is impractical to assume that a particular 'x' samples(values) makes a pulse signal.So, I need a method that can detect/distinguish each pulse signal depending on some parameters and save each pulse signal data.

The incoming pulse signals looks like the below

Answer 1

[Image Analyst]

Use findpeaks() in the Signal Processing Toolbox, or this http://www.mathworks.com/matlabcentral/fileexchange/25500-peakfinder

[Gova ReDDy]

Yes,the peaks of pulse signals are found using the function from above link.

But in realtime each time 500 samples(each collected at 2ms(500Hz) at microcontroller) will be sent to matlab respresenting 1 sec duration of pulse signal.So,in the below image we can see that out of 500 samples only around 412 values represent a pulse so these 412 values should be saved into some variable say P1 and the remaining 88 values should be added to the length of P2 next consecutive pulse signal detected. In this the first four pulse signals are to be saved in some variables so that they can be used to average with the currently detected pulse signal.

Can I know how to implement this or can you suggest any other idea to implement this kind of periodic moving average.

Answer 2

[Gova ReDDy]

I just improved the analog circuitry such that the output pulse signal(sampled at 500Hz) has a minimum value of zero whereas the previous circuit was resulting the pulse signal shown above where the minimum value was varying a lot.

Can someone suggest a method to divide this incoming pulse signal into periods such that each period contains a pulse signal.I also guess that each divided period(of pulse signal)doesn't contains same samples for which afterwards the interpolation will be used for fitting the samples between them.

I am not sure of what parameters can be used to distinguish and divide each pulse signal.

[Image Analyst]

What exactly are you looking for? What does "distinguish" mean to you? What kind of variables containing what kind of numerical values are you looking for?

[Gova ReDDy]

I want to implement the method mentioned here but the mentioned one as AC pulse signals I mean with different minimum values but whereas my pulse signals has a minimum value as zero which is confusing me to apply the segmentation point and also other thing is the first period selection or division.

My main idea is to average the current pulse signal with the last four pulse signals so that even if the current obtained pulse is not in good shape or amplitude(I mean when there is some disturbance with the measured pulse) can make it better(or similar like the one without disturbance).

Can you please suggest any solution for this.

[Image Analyst]

That didn't really answer the question. I was looking for something like "I want to find the starting and ending index of each hump" or "I want to find the index of the middle of the flat part between the humps." So just threshold

binarySignal = signal > 0;

Then use diff() to find the [0, 1] or [1, 0] transition indexes.

[Gova ReDDy]

Yes,I want to find the starting index of one hump and save the values(amplitudes) in a variable untill it encounters another starting index of next hump as shown below (The saved values correspond to a pulse signal)

I am confused of to solve when the 1 second duration samples has two half's of pulses corresponding the two different pulse signals. How to save the all values corresponding to a particular pulse signal(in which some them were present in the last 1 second duration samples and some in the current 1 sec duration samples).

As shown in the image below(obtained in labview) and will be similar to the matlab as well sampled at 500 hz.

[Gova ReDDy]

can you try to help me out with this .

[Image Analyst]

I thought I did.

[Gova ReDDy]

yes,it is working perfectly and is making the pulse signal into square signal.

But can I know how to

find the starting index(where the amplitude >some x value say 0.08) of one hump and save the values(amplitudes) in a variable untill it encounters another starting index of next hump(again when the amplitude exceeds 0.08) for every incoming 500 samples. as sometimes the 500 samples will have two half pulse signals as shown in my last post image.How to add the last half samples(in last 500 samples) of pulse signal to the current half samples to make it full pulse. Also how to save the first detected 4 pulse signals into some variables so that I will use this for averaging the currently detected pulse signal.

The signal data is attached.I hope you will unfold this implementation.