Filter design: MATLAB firls (Least-squares linear-phase FIR filter design) alternative?

Specific Domains Signal and Image Processing
8

Thank you for the code and information. I checked out the documentation, and did the following to reproduce something similar to the MATLAB code I was using.

center_freq = 20 # in Hz
filter_frequency_spread  = 6 # Hz +/- the center frequency
transition_width = 0.2
ffrequencies   = [ 
    0, 
    (1-transition_width)*(center_freq-filter_frequency_spread), 
    (center_freq-filter_frequency_spread),
    (center_freq+filter_frequency_spread), 
    (1+transition_width)*(center_freq+filter_frequency_spread), 
    nyquist, 
] /nyquist
idealresponse  = [ 0, 0, 1, 1, 0, 0, ]

# Order hard set to 200 as in MATLAB code
filterweights = remez(
    200, 
    [
        (ffrequencies[1], ffrequencies[2]) => 0, 
        (ffrequencies[3], ffrequencies[4]) => 1, 
        (ffrequencies[5], ffrequencies[6]) => 0 
    ], 
    Hz = EEG["srate"]/nyquist
)

And then to take a look at what’s going on with a fft

filterweights  = (firls(200,ffrequencies,idealresponse));

# compute its power spectrum
fft_filtkern  = abs.(fft(filterweights))
fft_filtkern  = fft_filtkern./maximum(fft_filtkern) # normalized to 1.0 for visual comparison ease

hz_filtkern   = range(0,nyquist, length=101); # list of frequencies in Hz corresponding to filter kernel

image
image770×289 11.1 KB

From the original MATLAB code:
image