There is a "weightingFilter" function that converts given signal to weighted signal, for A,B,C audio filters. Can we include our own filter in the function ?

The weighting filter I have has 44 frequencies and is taken from ISO-2631 standard. And my signal has 20 samples in discrete time between 0 to 10secs. The goal is to get weighted signal as output. If we can not include our filter in the MATLAB function, then how do we calculate weighted signal ?

[Jonas]

to which part of the iso are you referring? Is your sample rate enough, Most of the time the requirements for sampling rate are at 256 or 80 Hz, you have only 20/10s=2Hz

normally it is decribes how you have to apply your filter, e.g. a Direct form II Tranposed filter, including coefficients. FOr that example, you can e.g. use the Matlab function dfilt.df2t

[Walter Roberson]

80 Hz sounds too low to resolve more than 80/2 = 40 frequencies?

[Jonas]

yes, that's why i asked about the exact part of the iso

[Mario Malic]

You definitely need a lot more samples than that. Recommended time to capture the signal is 60 seconds, at least, or two mins if you have frequency components around 1 Hz (from Standard).

[Ishu Jaiswal]

I am referring ISO 2631-1:1997 Mechanical vibration and shock — Evaluation of human exposure to whole-body vibration — Part 1. In this filters frequency are separated by 1/3rd of octave. Specifically I am referring to W_d filter.

Can you please let me know, if I am planning a journey for 10 secs, then in order to increase the number of samples of input signals should I increase the sampling rate or plan a journey for longer duration of time ?

In my approach I am first converting the discrete signal to continuous using trapezoidal interpolation or spline. Then I am sampling the signal such that in frequency domain those frequencies are present in which filter has been designed (those 44 frequencies). Then multiplication is done between filter and FFT of signal. And at last converted to time domain.

What should be standard approach in this type of problem ? It's a little challenging for me right now due to little background.

[Ishu Jaiswal]

Also, the frequency components of the weighting filter ranges from 0.02Hz to 400Hz.

[Mario Malic]

I think the order should be:

1. Obtain the signal in the time domain
2. Filter the time signal for each 1/3 octave band, the signal is still in time domain (check octaveFilterBank function)
3. Make FFT, fix amplitude
4. Now you have signal in frequency domain, get the whole band, 0.01 Hz to 400 Hz. In code: filteredVelocitySpectrum(startRollOffFreqIdx : endRollOffFreqIdx)
5. For each band get RMS of signal, multiply it with the Wd value of according band
6. Then, get overall weighted value like written on the second line.

weightedVelRMS = rms(abs(filteredVelocitySpectrum(startRollOffFreqIdx : endRollOffFreqIdx, :)), 1) .* freqWeightsWmTable.FactorWv(startWvFreqIdx:endWvFreqIdx)';
ovWeightedVelRMS = sqrt(sum(weightedVelRMS.^2));

I hope it helps, maybe I have some mistakes in my understanding of the standard. However, they should make it easier to understand. There is some code for ISO 2631 on File Exchange, however comments have been stripped out and it's hard to figure out what's going on.

[Walter Roberson]

Taking a short signal with low time resolution and using an interpolating spline on it will not give you something that can be adequately analyzed at 44 frequencies. A sample rate of 2 Hz would require that the signal duration be... not sure, a couple of days?? in order to be able to resolve the frequencies. And that is assuming repeated signal.

You are probably going to have to increase your sampling rate

[Mario Malic]

Sampling rate definitely needs to be increased.