high-pass filter implementation in python

[saeedsab]

Hi
I want to subtract all the behaviour that have length scale bigger than data-length/10 from each vector. Here is the function I wrote:

def hpass(vec):
	cutoff_hz = len(vec)/10.
	sample_rate=len(vec)
	nyq_rate = sample_rate / 2.
	numtaps = 10
	warmup = numtaps - 1
	fir_coeff = firwin(numtaps, cutoff_hz/nyq_rate)
	delay = (warmup) #/ sample_rate
	t=np.arange(len(vec))
	t-=delay
	ln=len(t)-delay
	t=t[:ln]
	fl= lfilter(fir_coeff,1.0,vec)
	tmp=vec[:ln]-fl[t]
	filtered=(tmp)
        return filtered

but apparently this does not work probably since the delay part is not calculated right, can anybody help me with this?

Cheers

[YarrOfDoom]

It's been a while since I've done anything frequency related so could you explain what the value for delay should be?
Right now it is equal to warmup, since half of line 8 is commented out.

If you're using Python 2.x rather than 3.x the problem might be integer division related since warmup and sample_rate are both integer values (assuming you're using the full line 8 without anything commented out). Python 3.x integer division results in a float (so 3/2=1.5), but 2.x versions still have classic integer division where the results are rounded down (3/2=1).