Smoothing discrete data with hanning window

I am new to matlab analysis and I need to smooth dicharge times of a motor unit (discrete data points representing firing times). All previous literature uses "400 ms Hanning window" to smooth the firing rate (calculated from the discharge times). My question is, I have ,for example, an array containing 800 numbers (consecutive discharge times in sec) . How can I apply 400 ms Hanning window on this data? My sampling rate is 20K Hz.

[Adam]

You have nowhere near enough data to do this. 800 samples is only 4ms worth of data unless my on the spot maths is wrong.

[Star Strider]

See the fir1 documentation section on Multiband FIR Filter (link) for an example on how to design a FIR filter with a Hanning window.

Answer 1

[innavoig23]

I just created a repo on GitHub since I needed to do that too for my MSc thesis. Here's the link.

[Adam Danz]

+1 thanks for sharing, @Giovanni

Answer 2

[Adam Danz]

Here's a full description and examples.

https://www.mathworks.com/help/signal/ref/hann.html

[Gizem Yilmaz]

Thank you for your help, but I have already went through help pages. the problem is I have an array (1x800) which contains firing times in seconds. Even if I convert time into samples, data size do not change. When I create a Hann window of 400 ms, doesn't it mean that my window size L will be L=0.4*fsamp ? In my case, the samp.rate is 20k Hz and L will be 8000 points. How can I use this window on my 800 points data? I might be getting a concept wrong, if so, what might it be in terms of filtering?

[Adam Danz]

See Adam's comment under your question. I agree that if your sampling rate is 20Khz and you've got 800 samples, you simply don't have enough data to use a 400ms window.

20kHz is 0.05 milliseconds per sample. If you have 800 equispaced samples at that rate, that's only 40ms.

[Gizem Yilmaz]

I got this,however, my problem is that my data is not "800 equispaced samples". In my discharge times array I have indepented discharge times in sec, for example [16.75 16.9 18.12 .... 200.54]. I need to follow Hann windwing approach. I don't know how to pass my discrete data through hann window...

[Adam]

If that is the case how can you have a single sample rate? A scalar sample rate implies all data is the same distance apart on a uniform grid.

[Adam Danz]

@Gizem, If I'm interpreting your description correctly, you've sampled a motor unit at 20kHz and recorded the timestamps of each action potential. For simplicity, let's say the sampling rate was at 1ms intervals (1kHz), if my timestamps were

[2 5 7 10]

then I could replace that with a vector sampled a 1ms resolution where a 1 indicates a spike.

[0 1 0 0 1 0 1 0 0 1]

There is a 1 (a spike) at 2ms, 5ms, 7ms, and 10ms.

Is that an accurate description of your data?