def genFilters():
bp = [ ]
bp.append(ss.remez(50, [0, 0.02, 0.05, 0.5], [1,0]))
bp.append(ss.remez(50, [0, 0.02, 0.05, 0.20, 0.25, 0.5], [0, 1, 0]))
bp.append(ss.remez(50, [0, 0.20, 0.25, 0.5], [0, 1],
type = "hilbert"))
return bp
def HPman(self, fil_dict):
self._get_params(fil_dict)
if (self.N % 2 == 0): # even order, use odd symmetry (type III)
self._save(fil_dict,
sig.remez(self.N,[0, self.F_SB, self.F_PB, 0.5], [0, 1],
weight = [fil_dict['W_SB'],fil_dict['W_PB']], Hz = 1,
type = 'hilbert', grid_density = self.grid_density))
else: # odd order,
self._save(fil_dict,
sig.remez(self.N,[0, self.F_SB, self.F_PB, 0.5], [0, 1],
weight = [fil_dict['W_SB'],fil_dict['W_PB']], Hz = 1,
type = 'bandpass', grid_density = self.grid_density))
def filter_design():
if not glob.glob('ECoG_filter.h5'):
from scipy.signal import remez
f1 = remez(121, [0, 8, 12, 20, 24, 500], [0, 1, 0], Hz=1000)
f2 = remez(121, [0, 50, 60, 100, 110, 500], [0, 1, 0], Hz=1000)
f3 = remez(121, [0, 90, 100, 200, 210, 500], [0, 1, 0], Hz=1000)
filters = [f1, f2, f3]
with h5py.File('ECoG_filter.h5', 'w') as f:
f.create_dataset('filters', data=filters)
else:
with h5py.File('ECoG_filter.h5', 'r') as f:
filters = f['filters'][:]
return filters
def HPmin(self, fil_dict):
self._get_params(fil_dict)
(self.N, F, A, W) = remezord([self.F_SB, self.F_PB], [0, 1],
[self.A_SB, self.A_PB], Hz = 1, alg = self.alg)
# self.N = ceil_odd(N) # enforce odd order
fil_dict['W_SB'] = W[0]
fil_dict['W_PB'] = W[1]
if (self.N % 2 == 0): # even order
self._save(fil_dict, sig.remez(self.N, F,[0, 1], weight = W, Hz = 1,
type = 'hilbert', grid_density = self.grid_density))
else:
self._save(fil_dict, sig.remez(self.N, F,[0, 1], weight = W, Hz = 1,
type = 'bandpass', grid_density = self.grid_density))
def HPmin(self, fil_dict):
self._get_params(fil_dict)
(self.N, F, A, W) = remezord([self.F_SB, self.F_PB], [0, 1],
[self.A_SB, self.A_PB], fs = 1, alg = self.alg)
if not self._test_N():
return -1
# self.N = ceil_odd(N) # enforce odd order
fil_dict['W_SB'] = W[0]
fil_dict['W_PB'] = W[1]
if (self.N % 2 == 0): # even order
self._save(fil_dict, sig.remez(self.N, F,[0, 1], weight = W, fs = 1,
type = 'hilbert', grid_density = self.grid_density))
else:
self._save(fil_dict, sig.remez(self.N, F,[0, 1], weight = W, fs = 1,
type = 'bandpass', grid_density = self.grid_density))
def fir_remez_bpf(f_stop1,
f_pass1,
f_pass2,
f_stop2,
d_pass,
d_stop,
fs=1.0,
n_bump=5,
status=True):
"""
Design an FIR bandpass filter using remez with order
determination. The filter order is determined based on
f_stop1 Hz, f_pass1 Hz, f_pass2 Hz, f_stop2 Hz, and the
desired passband ripple d_pass dB and stopband attenuation
d_stop dB all relative to a sampling rate of fs Hz.
Mark Wickert October 2016, updated October 2018
"""
n, ff, aa, wts = bandpass_order(f_stop1,
f_pass1,
f_pass2,
f_stop2,
d_pass,
d_stop,
fsamp=fs)
# Bump up the order by N_bump to bring down the final d_pass & d_stop
N_taps = n
N_taps += n_bump
b = signal.remez(N_taps, ff, aa[0::2], wts, Hz=2)
if status:
log.info('Remez filter taps = %d.' % N_taps)
return b
def high_pass_filter(signal_input,
cut_off_freq,
sampling_freq,
filter_order=2):
"""Applying high-pass filter
Args:
signal_input: the input data (1d-array).
cut_off_freq: frequency below which the signals are attenuated.
sampling_freq: the sampling frequency of the recorded data.
filter_order: a number showing the complexity of the filter structure
`remez` function get the number of taps as input. The number of taps is the
number of terms in the filter, or the filter order plus one.
Returns:
filtered_signal: the filtered signal (1d-array)
filter_coefficients: the coefficients of the optimal filter for the high pass filter
"""
filter_coefficients = remez(
filter_order + 1,
[0, 0.9 * cut_off_freq, 1.1 * cut_off_freq, 0.5 * sampling_freq],
[0, 1.0],
Hz=sampling_freq)
filtered_signal = lfilter(filter_coefficients, 1, signal_input)
return filtered_signal, filter_coefficients
def coeffbandpass_filter(lcut1, lcut2, hcut1, hcut2, fs, n):
"""
Method used to design a bandpass filter
Parameters:
data : eeg signal to filter
*__________*
/| |\
/ | | \
/ | | \
___*/ | | \*____
lcut1 lcut2 hcut1 hcut2
lowcut1, lowcut2, highcut1, highcut2 : frequency band of the fir filter
fs : sampling frequency
n : number of iteration for the Remez algorithm
"""
low1 = lcut1 / fs
low2 = lcut2 / fs
high1 = hcut1 / fs
high2 = hcut2 / fs
b = signal.remez(n, [0, low1, low2, high1, high2, 0.5], [0, 1, 0])
#print(b)
"""freq, response = signal.freqz(b)
ampl = np.abs(response)
fig1 = plt.figure()
ax1 = fig1.add_subplot(111)
ax1.semilogy(fs*freq/(2*np.pi), ampl, 'b-') # freq in Hz"""
return b
def bpf_design(fs: int, fcutoff: float, flow: float=300., L: int=256):
"""
Design FIR bandpass filter coefficients "b"
fcutoff: cutoff frequency [Hz]
fs: sampling frequency [Hz]
flow: low cutoff freq [Hz] to eliminate rumble or beating carriers
L: number of taps (more taps->narrower transition band->more CPU)
https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.remez.html
"""
firtype = 'firwin'
if firtype == 'remez':
# 0.8*fc is arbitrary, for finite transition width
b = signal.remez(L, [0, 0.8*flow,
flow, 0.8*fcutoff,
fcutoff, 0.5*fs],
[0., 1., 0.], Hz=fs)
elif firtype == 'firwin':
b = signal.firwin(L, [flow, fcutoff], pass_zero=False, width=100, nyq=0.5*fs,
window='kaiser', scale=True)
elif firtype == 'matlab':
assert L % 2 != 0, 'must have odd number of taps'
from oct2py import Oct2Py
with Oct2Py() as oc:
oc.eval('pkg load signal')
b = oc.fir1(L+1, [0.03, 0.35], 'bandpass')
return b
def Bandpass(lowcut, highcut, trans_width, fs):
numtaps = 100
edges = [0, lowcut - trans_width,
lowcut, highcut,
highcut + trans_width, 0.5*fs]
taps = sig.remez(numtaps, edges, [0, 1, 0], Hz=fs)
return taps
def lowpass_remez(fl1, fl2, x, map=2, mas=40, Fs=1000.0):
"""Simple interface for lowpass-filter using Parks-McClellan algorithm.
:Parameters:
fl1:
float, lower edge-frequency for transition band
fl2:
float, upper edge-frequency for transition band
x:
array, data to filter. Can be array of arbitrary dimension. The first index must
be the time index.
map:
float, maximum attenuation in pass band
mas:
float, minimum attenuation in stop band
Fs:
float, Sampling-frequency of the data
:Returns:
y: array, The filtered data
"""
map = float(-abs(map))
mas = float(-abs(mas))
n, f, a, w = remezord([fl1, fl2], [1, 0],
[1 - (10**(map / 10)), 10**(mas / 10)],
Hz=Fs)
b = remez(n, f, a, w)
#z=lfilter(b,a,xn)
y = filtfilt(b, [1], x)
return y
def test_firpm_1(self):
x = [i for i in range(len(self.f1))]
ipf = ip.interp1d(x, self.f1)
f1_new = ipf(np.linspace(x[0], x[-1], 2 * len(x)))
FIR = FIRDesign.firpm(self.n1, self.f1, self.a1)
fir = signal.remez(self.n1 + 1, f1_new, self.a1, fs=2)
self.assertTrue(np.all(FIR[0] == fir))
def FilterAndDownSample(x, fs):
"""
Parametros:
x: Señal a filtrar y
fs: Frecuencia de muestreo
"""
# traps es el número de términos en el filtro, o el orden de filtro más uno.
n_taps = 50
# Calcule el filtro óptimo minimax utilizando el algoritmo de intercambio Remez.
coef = remez(n_taps, [0, F_BW, F_BW * 1.4, fs / 2], [1, 0],
Hz=fs) # (*@ \label{code:remezFunction} @*)
x_filter = lfilter(coef, 1.0, x)
if PLOT:
PlotFilterCharacteristic(coef, fs)
dec_rate = int(fs / F_BW)
x_downsample = x_filter[0::dec_rate]
# Se calcula la nueva frecuencia de muestreo
fs_y = fs / dec_rate
if PLOT:
PlotSpectrum(x_downsample, "x downsample", "x_downsample_spectrum.pdf",
fs_y)
PlotConstelation(x_downsample, "x downsample",
"x_downsample_constelation.pdf")
return x_downsample, fs_y
def design(self, params):
coeffs = signal.remez(**params)
den = np.zeros_like(coeffs)
den[0] = 1
return coeffs, den
def example_hilbert_robust_remez():
'''
1. use the Remez exchange algorithm to design a lowpass filter
2. modulation thee lowpass impulse-response by a complex sinusoid at frequency fs/4
'''
count = 257
fs = 22050
fn = fs / 2 # nyquest
f1 = 530 # transition bandwidth
f2 = fn - f1 # upper transition bandwidth
count = 257
bands = [0, f2 - fs / 4, fs / 4, fn]
lpfir = signal.remez(count, bands, [1, 0], [1, 10], fs=fs)
# modulate lowpass to single-sideband
#csin = np.array([1j])
modsin = np.power(np.full((count), 1j), np.arange(count))
fir = lpfir * modsin
response = np.fft.fft(fir, 4096)
gain = np.abs(response)
gain_db = 20 * np.log10(gain)
gain_db = gain_db - np.nanmax(gain_db)
plt.plot(np.fft.fftshift(gain_db))
plt.show()
def filter_examples():
'''
Plots some example bandpass filters with different numbers of taps
'''
bands = [0, .19, .21, .29, .31, .5]
gains = [0, 1, 0]
# bands = [0, .19, .2, .3, .31, .5]
# gains = [0, 0, 1, 1, 0, 0]
B = []
a = 1
taps = range(10, 131, 30)
for t in taps:
B.append(remez(t, bands, gains, type = 'bandpass', maxiter = 1000,
grid_density = 32))
# B.append(firwin2(t, bands, gains, nyq = 0.5))
fig = plt.figure()
ax1 = fig.add_subplot(1,1,1)
ax1.set_title('Frequency Responses for Various FIR lengths')
ax1.set_ylabel('Amplitude [dBV]')
ax1.set_xlabel('Frequency [Rad/Sample]')
for i, b in enumerate(B):
w, h = freqz(b, a)
ax1.plot(w, 20*log10(abs(h)), label = '%d taps' % taps[i])
ax1.legend()
fig.show()
raw_input('Continue?...')
return
def fir_optim(Fp, ap, Fa, aa, graf=False):
"""
Donats una serie de paràmetres, retorna un array de coeficients b que formen el filtre amb les especificacions donades.
\t Fp: frequencia discreta limit de la banda de pas
\t ap: arrissat en dB de la banda de pas
\t Fa: frequencia discreta limit de la banda de rebuig
\t aa: minima atenuacio en dB de la banda de rebuig respecte de la de pas
\t graf: representació del filtre creat. El seu valor pot ser True o False.[argument opcional]
\n
"""
dp = (10**(ap / 10) - 1) / (10**(ap / 20) + 1)
da = 10**(-aa / 20)
L = 2
while True:
b = scs.remez(
L, [0, Fp, Fa, 0.5], [1, 0], [da, dp]
) # Anem calculant conjunts de coeficients de ordre creixent fins a retornar.
if compleix(
b, Fp, dp, Fa, da
): # Comprovem si aquells coeficients compleixen els requisits especificats per el filtre.
if graf == True:
representa_FIR(
b, Fp, ap, Fa,
aa) # Si es compleix, retornem i el bucle acaba.
return b
L += 1
def do_plot(self):
recv_f0 = self.params.get('recv_f0', 2250)
recv_f1 = self.params.get('recv_f1', 3150)
recv_f = [recv_f0, recv_f1]
sample_f = 48000 # sampling rate, Hz, must be integer
plt.figure()
plt.ylim(-60, 5)
plt.xlim(0, 5500)
plt.grid(True)
plt.xlabel('Frequency (Hz)')
plt.ylabel('Gain (dB)')
plt.title('{}Hz, {}Hz'.format(recv_f[0], recv_f[1]))
filters = []
fbw = [(recv_f[1] - recv_f[0]) * 0.85, (recv_f[1] - recv_f[0]) * 0.8]
for i in range(2):
f = recv_f[i]
filter_bp = signal.remez(
80, [0, f - fbw[i], f, f, f + fbw[i], sample_f / 2], [0, 1, 0],
fs=sample_f,
maxiter=100)
filters.append(filter_bp)
w, h = signal.freqz(filters[i], [1], worN=2500)
plt.plot(0.5 * sample_f * w / np.pi, 20 * np.log10(np.abs(h)))
plt.plot((f, f), (10, -60), color='red', linestyle='dashed')
plt.plot((500, 500), (10, -60), color='blue', linestyle='dashed')
plt.plot((700, 700), (10, -60), color='blue', linestyle='dashed')
plt.show()
def band_pass_filter(signal_input,
lower_bound_freq,
higher_bound_freq,
sampling_freq,
filter_order=2):
"""Applying band-pass filter
Args:
signal_input: the input data (1d-array).
lower_bound_freq: the lower bound frequency below which the signal must be completely attenuated.
higher_bound_freq: the higher bound frequency above which the signal must be completely attenuated.
sampling_freq: the sampling frequency of the recorded data.
filter_order: a number showing the complexity of the filter structure
`remez` function get the number of taps as input. The number of taps is the
number of terms in the filter, or the filter order plus one.
Returns:
filtered_signal: the filtered signal (1d-array)
filter_coefficients: the coefficients of the optimal filter for the band pass filter
"""
filter_coefficients = remez(filter_order + 1, [
0, 0.9 * lower_bound_freq, lower_bound_freq, higher_bound_freq,
1.1 * higher_bound_freq, 0.5 * sampling_freq
], [0, 1.0, 0],
Hz=sampling_freq)
filtered_signal = lfilter(filter_coefficients, 1, signal_input)
return filtered_signal, filter_coefficients
def _design_remez(self, maxiter=25):
print("Taps = ", self.taps)
print("Freqs = ", self.freqs)
print("Gains = ", self.gains)
self.B = signal.remez(self.taps, self.freqs,
self.gains, maxiter=maxiter,
Hz=self.sample_rate)
def lab6_ex3():
# set parameters of system
fs = 2000 #sampling frequency (Hz)
fn = fs/2 #nyqvist frequency
fc = array([0,100,200,400,500,1000]) #corner frequencies
fb = array([0,1.0,0]) #desired band gains for each pair of corner frequencies
deg = 46 #degree of polynomial
nt = deg + 1 #number of taps = degree + 1
a = array([1.0,0]) #(no feedback (FIR), poles at origin reduced to 1 for simplification)
# create remez filter, frequency response, and plot
b = remez(nt,fc,fb,Hz=fs)
w,h = freqz(b,a)
plot(w/pi*fn, 20*log10(abs(h)),'b-')
title('FIR of given frequency response using remez algo')
xlabel('frequency')
ylabel('magnitude (dB scale)')
grid()
show()
print('\nFilter requires 47 taps for stop-band attenuation of at least -40 dB')
z,p,k = tf2zpk(b,a)
zplane(z,p)
title('zplane of remez-FIR of given frequency response (47 taps)')
show()
def filter_data(self):
'''Creates a low pass filter to filter out high frequency noise'''
lpf = remez(self.bin_numbers, [0, 0.1, 0.25, 0.5], [1.0, 0.0])
filt_data = np.zeros(self.num_sensors+self.num_sharp_sensors)
for i in range(self.num_sensors+self.num_sharp_sensors):
filt_data[i] = np.dot(lpf, self.raw_dat_array[:,i])
return filt_data
def lowpass_remez(fl1,fl2,x,map=2,mas=40,Fs=1000.0):
"""Simple interface for lowpass-filter using Parks-McClellan algorithm.
:Parameters:
fl1:
float, lower edge-frequency for transition band
fl2:
float, upper edge-frequency for transition band
x:
array, data to filter. Can be array of arbitrary dimension. The first index must
be the time index.
map:
float, maximum attenuation in pass band
mas:
float, minimum attenuation in stop band
Fs:
float, Sampling-frequency of the data
:Returns:
y: array, The filtered data
"""
map = float(-abs(map))
mas = float(-abs(mas))
n,f,a,w = remezord([fl1,fl2],[1,0],[1-(10**(map/10)),10**(mas/10)],Hz=Fs)
b=remez(n,f,a,w)
#z=lfilter(b,a,xn)
y=filtfilt(b,[1],x)
return y
def get_filtered(self, array, band):
processed = np.array(array)
processed = signal.detrend(processed)
processed = utils.butter_bandpass_filter(processed,
0.3,
2.5,
self.fps,
order=3)
signal_line = processed - smooth(processed, 5)
signal_line = utils.butter_bandpass_filter(signal_line,
0.05,
0.15,
self.fps,
order=9)
analytic_signal = signal.hilbert(signal_line)
high = np.mean(np.abs(analytic_signal))
low = np.mean(low_envelope(signal_line))
d_signal = signal.detrend(signal_line)
removed = filter(lambda x: low < x < high, d_signal)
edges = [0, band[0], band[1], band[2], band[3], 0.5 * self.fps]
taps = signal.remez(3, edges, [0, 1, 0], Hz=self.fps, grid_density=20)
filttered_color = np.convolve(taps, removed)[len(taps) // 2:]
return filttered_color
def generate_lowpass(self, fs, cutoff, trans_width, numtaps):
taps = signal.remez(numtaps,
[0, cutoff, cutoff + trans_width, 0.5 * fs],
[1, 0],
Hz=fs)
w, h = signal.freqz(taps, [1], worN=2000)
return w, h, taps
def __init__(self, fir_len, fir_bits, tb_width):
tb_ctr = 1/(2*4)
pass_corner = tb_ctr - (tb_ctr*tb_width/2)
stop_corner = tb_ctr + (tb_ctr*tb_width/2)
fir_bands = [0, pass_corner, stop_corner, 0.5]
b = signal.remez(fir_len, fir_bands, [1, 0])
coeff_scl = 2**(fir_bits-1)
self.fir_coeff = np.floor(b*coeff_scl + 0.5)
# Dump Coefficients?
if 1:
write_meminit("fir4dec_coeff.v", self.fir_coeff)
self.fir_coeff = self.fir_coeff/coeff_scl;
# plot FIR response?
if 1:
W, H = signal.freqz(self.fir_coeff)
plt.figure()
plt.plot(W/(2*np.pi), 20*np.log10(np.abs(H)))
plt.grid()
plt.xlabel("Freq (normalized)")
plt.ylabel("dB")
plt.title("fir4dec response (close to continue sim)")
plt.show()
def prototype_filter_remez():
""" ASSIGNMENT 2
Compute the prototype filter used in subband coding. The filter
is a 512-point lowpass FIR h[n] with bandwidth pi/64 and stopband
starting at pi/32
You should use the remez routine (signal.remez()). See
http://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.remez.html
"""
from scipy import signal
M = 512
Fs = 44100
F_nyquist = Fs // 2
Fpass = F_nyquist * (1 / 128)
Fstop = F_nyquist * (1 / 32)
bpass = signal.remez(numtaps=M,
bands=[0, Fpass, Fstop, F_nyquist],
desired=[2, 0],
Hz=Fs)
return bpass
def __init__(self, *args, **kwargs):
Modulator.__init__(self, *args, **kwargs)
self.taps = 81
self.kernel = remez(self.taps, [0.1, 0.4], [1.0], type='hilbert')
self.zi = np.zeros(len(self.kernel) - 1)
G = self.taps / 2
self.overlap = np.zeros(G)
def __init__(self, stationMHz):
self.sdr = rtlsdr.RtlSdr()
self.validsGains = self.sdr.get_gains()
self.indexGain = 10
self.f_offset = 250000 # Desplazamiento para capturar
self.f_station = stationMHz # Frecuencia de radio
self.dec_rate = int(FS / F_BW)
self.fs_y = FS / (self.dec_rate)
self.coef = remez(N_TRAPS, [0, F_BW, F_BW * 1.4, FS / 2], [1, 0],
Hz=FS)
self.expShift = (-1.0j * 2.0 * np.pi * self.f_offset / FS)
# Se configura los parametros
self.dec_audio = int(self.fs_y / AUDIO_FREC)
self.stream = sd.OutputStream(device='default',
samplerate=int(self.fs_y /
self.dec_audio),
channels=1,
dtype='float32')
self.beginListening = 0
self.soundQueue = queue.Queue()
self.samplesQueue = queue.Queue()
def fir_remez_bsf(f_pass1,
f_stop1,
f_stop2,
f_pass2,
d_pass,
d_stop,
fs=1.0,
N_bump=5):
"""
Design an FIR bandstop filter using remez with order
determination. The filter order is determined based on
f_pass1 Hz, f_stop1 Hz, f_stop2 Hz, f_pass2 Hz, and the
desired passband ripple d_pass dB and stopband attenuation
d_stop dB all relative to a sampling rate of fs Hz.
Mark Wickert October 2016
"""
n, ff, aa, wts = optfir.remezord([f_pass1, f_stop1, f_stop2, f_pass2], [
1, 0, 1
], [1 - 10**(-d_pass / 20.), 10**(-d_stop / 20.), 1 - 10**(-d_pass / 20.)],
fsamp=fs)
# Bump up the order by N_bump to bring down the final d_pass & d_stop
N_taps = n
N_taps += N_bump
b = signal.remez(N_taps, ff, aa[0::2], wts, Hz=2)
print('Remez filter taps = %d.' % N_taps)
return b
def BSman(self, fil_dict):
self._get_params(fil_dict)
self.N = round_odd(self.N) # enforce odd order
self._save(fil_dict, sig.remez(self.N,[0, self.F_PB, self.F_SB,
self.F_SB2, self.F_PB2, 0.5],[1, 0, 1],
weight = [fil_dict['W_PB'],fil_dict['W_SB'], fil_dict['W_PB2']],
Hz = 1, grid_density = self.grid_density))
def BPman(self, fil_dict):
self._get_params(fil_dict)
self._save(fil_dict,
sig.remez(self.N,[0, self.F_SB, self.F_PB,
self.F_PB2, self.F_SB2, 0.5],[0, 1, 0],
weight = [fil_dict['W_SB'],fil_dict['W_PB'], fil_dict['W_SB2']],
Hz = 1, grid_density = self.grid_density))
def test_hilbert(self):
N = 11 # number of taps in the filter
a = 0.1 # width of the transition band
# design an unity gain hilbert bandpass filter from w to 0.5-w
h = remez(11, [a, 0.5-a], [1], type='hilbert')
# make sure the filter has correct # of taps
assert_(len(h) == N, "Number of Taps")
# make sure it is type III (anti-symmetric tap coefficients)
assert_array_almost_equal(h[:(N-1)//2], -h[:-(N-1)//2-1:-1])
# Since the requested response is symmetric, all even coeffcients
# should be zero (or in this case really small)
assert_((abs(h[1::2]) < 1e-15).all(), "Even Coefficients Equal Zero")
# now check the frequency response
w, H = freqz(h, 1)
f = w/2/np.pi
Hmag = abs(H)
# should have a zero at 0 and pi (in this case close to zero)
assert_((Hmag[[0, -1]] < 0.02).all(), "Zero at zero and pi")
# check that the pass band is close to unity
idx = np.logical_and(f > a, f < 0.5-a)
assert_((abs(Hmag[idx] - 1) < 0.015).all(), "Pass Band Close To Unity")
def test_hilbert(self):
N = 11 # number of taps in the filter
a = 0.1 # width of the transition band
# design an unity gain hilbert bandpass filter from w to 0.5-w
h = remez(11, [a, 0.5-a], [1], type='hilbert')
# make sure the filter has correct # of taps
assert_(len(h) == N, "Number of Taps")
# make sure it is type III (anti-symmetric tap coefficients)
assert_array_almost_equal(h[:(N-1)//2], -h[:-(N-1)//2-1:-1])
# Since the requested response is symmetric, all even coeffcients
# should be zero (or in this case really small)
assert_((abs(h[1::2]) < 1e-15).all(), "Even Coefficients Equal Zero")
# now check the frequency response
w, H = freqz(h, 1)
f = w/2/np.pi
Hmag = abs(H)
# should have a zero at 0 and pi (in this case close to zero)
assert_((Hmag[[0, -1]] < 0.02).all(), "Zero at zero and pi")
# check that the pass band is close to unity
idx = np.logical_and(f > a, f < 0.5-a)
assert_((abs(Hmag[idx] - 1) < 0.015).all(), "Pass Band Close To Unity")
def _recv_decode_init_FIR(self, recv_f):
self._filters = []
fbw = [(recv_f[1] - recv_f[0]) * 0.85, (recv_f[1] - recv_f[0]) * 0.8]
for i in range(2):
f = recv_f[i]
filter_bp = signal.remez(
80, [0, f - fbw[i], f, f, f + fbw[i], sample_f / 2], [0, 1, 0],
fs=sample_f,
maxiter=100)
self._filters.append(filter_bp)
if not plot_spectrum:
return
import matplotlib.pyplot as plt
plt.figure()
plt.ylim(-60, 5)
plt.xlim(0, 5500)
plt.grid(True)
plt.xlabel('Frequency (Hz)')
plt.ylabel('Gain (dB)')
plt.title('{}Hz, {}Hz'.format(recv_f[0], recv_f[1]))
fbw = [(recv_f[1] - recv_f[0]) * 0.85, (recv_f[1] - recv_f[0]) * 0.8]
for i in range(2):
f = recv_f[i]
w, h = signal.freqz(self._filters[i], [1], worN=2500)
plt.plot(0.5 * sample_f * w / np.pi, 20 * np.log10(np.abs(h)))
plt.plot((f, f), (10, -100), color='red', linestyle='dashed')
plt.plot((500, 500), (10, -100), color='blue', linestyle='dashed')
plt.plot((700, 700), (10, -100), color='blue', linestyle='dashed')
plt.show()
def __init__(self, fir_len, fir_bits, tb_width):
tb_ctr = 1 / (2 * 8)
pass_corner = tb_ctr - (tb_ctr * tb_width / 2)
stop_corner = tb_ctr + (tb_ctr * tb_width / 2)
fir_bands = [0, pass_corner, stop_corner, 0.5]
b = signal.remez(fir_len, fir_bands, [1, 0])
coeff_scl = 2**(fir_bits - 1)
self.fir_coeff = np.floor(b * coeff_scl + 0.5)
# Dump Coefficients?
if 0:
LUT = np.zeros(256, dtype=np.int)
LUT[0:fir_len] = self.fir_coeff
write_memh("fir8dec_coeff.memh", LUT)
self.fir_coeff = self.fir_coeff / coeff_scl
# plot FIR response?
if 1:
W, H = signal.freqz(self.fir_coeff)
plt.figure()
plt.plot(W / (2 * np.pi), 20 * np.log10(np.abs(H)))
plt.grid()
plt.xlabel("Freq (normalized)")
plt.ylabel("dB")
plt.title("fir8dec response (close to continue sim)")
plt.show()
def prototype_filter():
"""
Computes the prototype filter used in subband coding. The filter
is a 512-point lowpass FIR h[n] with bandwidth pi/64 and stopband
starting at pi/32
"""
# number of lowpass points
lowpass_points = 512
#setting sampling frequency
fs = np.pi
#pass frequency
pass_frequency = fs / 128
#stop frequency
stop_frequency = fs / 32
#filter
filter = signal.remez(numtaps=lowpass_points,
bands=[0, pass_frequency, stop_frequency, fs],
desired=[2, 0],
fs=2 * fs)
return filter
def main():
# Compute filter coefficients with SciPy.
coef = signal.remez(80, [0, 0.1, 0.1, 0.5], [1, 0])
fir = FIR(coef)
# Simulate for different frequencies and concatenate
# the results.
in_signals = []
out_signals = []
for frequency in [0.05, 0.07, 0.1, 0.15, 0.2]:
tb = TB(fir, frequency)
fragment = autofragment.from_local()
sim = Simulator(fragment, Runner())
sim.run(100)
in_signals += tb.inputs
out_signals += tb.outputs
# Plot data from the input and output waveforms.
plt.plot(in_signals)
plt.plot(out_signals)
plt.show()
# Print the Verilog source for the filter.
print(verilog.convert(fir.get_fragment(),
ios={fir.i, fir.o}))
def fir_calc_filter(Fs, Fpb, Fsb, Apb, Asb, N):
bands = np.array([0., Fpb/Fs, Fsb/Fs, .5])
# Remez weight calculation:
# https://www.dsprelated.com/showcode/209.php
err_pb = (1 - 10**(-Apb/20))/2 # /2 is not part of the article above, but makes it work much better.
err_sb = 10**(-Asb/20)
w_pb = 1/err_pb
w_sb = 1/err_sb
h = signal.remez(
N+1, # Desired number of taps
bands, # All the band inflection points
[1,0], # Desired gain for each of the bands: 1 in the pass band, 0 in the stop band
[w_pb, w_sb]
)
(w,H) = signal.freqz(h)
Hpb_min = min(np.abs(H[0:int(Fpb/Fs*2 * len(H))]))
Hpb_max = max(np.abs(H[0:int(Fpb/Fs*2 * len(H))]))
Rpb = 1 - (Hpb_max - Hpb_min)
Hsb_max = max(np.abs(H[int(Fsb/Fs*2 * len(H)+1):len(H)]))
Rsb = Hsb_max
print("Rpb: %fdB" % (-dB20(Rpb)))
print("Rsb: %fdB" % -dB20(Rsb))
return (h, w, H, Rpb, Rsb, Hpb_min, Hpb_max, Hsb_max)
def fir_remez_hpf(f_stop, f_pass, d_pass, d_stop, fs = 1.0, N_bump=5):
"""
Design an FIR highpass filter using remez with order
determination. The filter order is determined based on
f_pass Hz, fstop Hz, and the desired passband ripple
d_pass dB and stopband attenuation d_stop dB all
relative to a sampling rate of fs Hz.
Mark Wickert October 2016
"""
# Transform HPF critical frequencies to lowpass equivalent
f_pass_eq = fs/2. - f_pass
f_stop_eq = fs/2. - f_stop
# Design LPF equivalent
n, ff, aa, wts=optfir.remezord([f_pass_eq,f_stop_eq], [1,0],
[1-10**(-d_pass/20.),10**(-d_stop/20.)],
fsamp=fs)
# Bump up the order by N_bump to bring down the final d_pass & d_stop
N_taps = n
N_taps += N_bump
b = signal.remez(N_taps, ff, aa[0::2], wts,Hz=2)
# Transform LPF equivalent to HPF
n = np.arange(len(b))
b *= (-1)**n
print('Remez filter taps = %d.' % N_taps)
return b
def better_envelope(rf_in):
# cutoff == 0.2 and 0.8 from fs = 20.832 MHz
# f_cutoff = 0.2*(fs/2) = 2.0832MHz, 0.8*(fs/2) = 8.3328,
# f0 = fs/4 = 5.208MHz fixed in Verasonics US systems
cutoff_low = 0.2
cutoff_high = 0.8
# fs = 20.832
num_taps = 10
# This is equivalent to B=firpm(10,[.2 .8],[1 1],'Hilbert');
# hilbert is a 90-phase transform.
coefficients = remez(num_taps + 1, [cutoff_low / 2, cutoff_high / 2], [1],
type='hilbert')
Q = filtfilt(coefficients, 1, rf_in) # zero-phase
envelope = np_sqrt(rf_in**2 + Q**2)
b, a = butter(5, 0.25, btype='low')
envelope_filtered = filtfilt(b,
a,
envelope,
axis=0,
padtype='odd',
padlen=3 * (max(len(b), len(a)) - 1))
envelope_filtered.clip(min=0, out=envelope_filtered)
return envelope_filtered
def DIFFman(self, fil_dict):
self._get_params(fil_dict)
if not self._test_N():
return -1
self.N = ceil_even(self.N) # enforce even order
self._save(fil_dict, sig.remez(self.N,[0, self.F_PB],[np.pi*fil_dict['W_PB']],
Hz = 1, type = 'differentiator', grid_density = self.grid_density))
def LPmin(self, fil_dict):
self._get_params(fil_dict)
(self.N, F, A, W) = remezord([self.F_PB, self.F_SB], [1, 0],
[self.A_PB, self.A_SB], Hz = 1, alg = self.alg)
fil_dict['W_PB'] = W[0]
fil_dict['W_SB'] = W[1]
self._save(fil_dict, sig.remez(self.N, F, [1, 0], weight = W, Hz = 1,
grid_density = self.grid_density))
def test_compare(self):
# test comparison to MATLAB
k = [0.024590270518440, -0.041314581814658, -0.075943803756711,
-0.003530911231040, 0.193140296954975, 0.373400753484939,
0.373400753484939, 0.193140296954975, -0.003530911231040,
-0.075943803756711, -0.041314581814658, 0.024590270518440]
h = remez(12, [0, 0.3, 0.5, 1], [1, 0], Hz=2.)
assert_allclose(h, k)
h = [-0.038976016082299, 0.018704846485491, -0.014644062687875,
0.002879152556419, 0.016849978528150, -0.043276706138248,
0.073641298245579, -0.103908158578635, 0.129770906801075,
-0.147163447297124, 0.153302248456347, -0.147163447297124,
0.129770906801075, -0.103908158578635, 0.073641298245579,
-0.043276706138248, 0.016849978528150, 0.002879152556419,
-0.014644062687875, 0.018704846485491, -0.038976016082299]
assert_allclose(remez(21, [0, 0.8, 0.9, 1], [0, 1], Hz=2.), h)
def LPman(self, fil_dict):
self._get_params(fil_dict)
if not self._test_N():
return -1
self._save(fil_dict,
sig.remez(self.N,[0, self.F_PB, self.F_SB, 0.5], [1, 0],
weight = [fil_dict['W_PB'],fil_dict['W_SB']], Hz = 1,
grid_density = self.grid_density))
def __init__(self, numtaps, bands, gains, fs, num_channels = 4):
self.numtaps = numtaps
self.fs = fs
self.num_channels = num_channels
self.b = signal.remez(numtaps, bands, gains, Hz = fs)[::-1]
self.x = np.zeros((num_channels,numtaps))
def HILman(self, fil_dict):
self._get_params(fil_dict)
if not self._test_N():
return -1
self._save(fil_dict, sig.remez(self.N,[0, self.F_SB, self.F_PB,
self.F_PB2, self.F_SB2, 0.5],[0, 1, 0],
weight = [fil_dict['W_SB'],fil_dict['W_PB'], fil_dict['W_SB2']],
Hz = 1, type = 'hilbert', grid_density = self.grid_density))
def _design_filter(self, FS):
if not FS in self._coefs_cache:
bands = [0, min(self.fs, self.fp), max(self.fs, self.fp), FS / 2]
gains = [int(self.fp < self.fs), int(self.fp > self.fs)]
b, a = signal.remez(self.order, bands, gains, Hz=FS), [1]
self._coefs_cache[FS] = (b, a)
else:
b, a = self._coefs_cache[FS]
return b, a
def __init__(self, controller='/autobed_height_controller'):
self.controller = controller
self.goal_pub = rospy.Publisher(controller+'/command', JointTrajectory)
self.state_sub = rospy.Subscriber(controller+'/state',
JointTrajectoryControllerState, self.state_cb)
self.joint_names = None
self.physical_sub = rospy.Subscriber('/abdout0', FloatArrayBare,
self.init_physical)
#Low pass filter design
self.bin_numbers = 5
self.bin_numbers_for_leg_filter = 21
self.collated_cal_angle = np.zeros((self.bin_numbers, 1))
self.collated_cal_angle_for_legs = np.zeros((
self.bin_numbers_for_leg_filter, 1))
self.filtered_angle = None
self.lpf = remez(self.bin_numbers, [0, 0.1, 0.25, 0.5], [1.0, 0.0])
self.lpf_for_legs = remez(self.bin_numbers_for_leg_filter,
[0, 0.0005, 0.1, 0.5], [1.0, 0.0])
def BPmin(self, fil_dict):
self.get_params(fil_dict)
(self.N, F, A, W) = pyfda_lib.remezord([self.F_SB, self.F_PB,
self.F_PB2, self.F_SB2], [0, 1, 0],
[self.A_SB, self.A_PB, self.A_SB2], Hz = 1, alg = self.alg)
fil_dict['W_SB'] = W[0]
fil_dict['W_PB'] = W[1]
fil_dict['W_SB2'] = W[2]
self.save(fil_dict, sig.remez(self.N,F,[0, 1, 0], weight = W, Hz = 1,
grid_density = self.grid_density))
def fircoef(L: int, fc: float, fs: int) -> np.ndarray:
"""
remez uses normalized frequency, where 0.5 is Nyquist frequency
fc: corner frequency [Hz]
"""
fcn = fc / (0.5*fs)
return signal.remez(L, [0, 0.6*fcn, fcn, 0.5], [0, 1])
def amfm_CCA(im):
"""
Channel component analysis for AM-FM
Input
- im : 1d vector that contains a time series
Output
- ia : Instantaneous amplitude computed for 3 channels
- ip : Instantaneous phase computed for 3 channels
- ifeq: Instantaneous frequency computed for 3 channels
"""
# Filter bank
bp = [ ]
# Low pass 0 0.02
bp.append(ss.remez(50, [0, 0.02,
0.05, 0.5],
[1,0]))
# Pass band 0.02 0.25
bp.append(ss.remez(50, [0, 0.02, 0.05,
0.20, 0.25, 0.5],
[0, 1, 0]))
# High pass 0.25 0.5
bp.append(ss.remez(50, [0, 0.20,
0.25, 0.5],
[0, 1],
type = "hilbert"))
# apply filterbank
filt = lambda x: ss.convolve(im,x,'same')
in_channels = map(filt,bp)
# compute IA, IP and IF from filterbank output
out_channels = map(qea,in_channels)
# Organize results into a matrix of channels by time points
ia = []
ip = []
ifeq = []
for chan in out_channels:
ia.append(chan[0])
ip.append(chan[1])
ifeq.append(chan[2])
ia = np.array(ia)
ip = np.array(ip)
ifeq = np.array(ifeq)
return(ia,ip,ifeq)
def filter_data(self):
'''Creates a low pass filter to filter out high frequency noise'''
lpf = remez(self.bin_numbers, [0, 0.05, 0.1, 0.5], [1.0, 0.0])
filt_data = np.zeros(self.num_sensors)
for i in xrange(self.num_sensors):
with self.frame_lock:
if np.shape(lpf) == np.shape(self.raw_dat_array[:,i]):
filt_data[i] = np.dot(lpf, self.raw_dat_array[:,i])
else:
pass
return filt_data
def BSmin(self, fil_dict):
self._get_params(fil_dict)
(N, F, A, W) = remezord([self.F_PB, self.F_SB,
self.F_SB2, self.F_PB2], [1, 0, 1],
[self.A_PB, self.A_SB, self.A_PB2], Hz = 1, alg = self.alg)
self.N = round_odd(N) # enforce odd order
fil_dict['W_PB'] = W[0]
fil_dict['W_SB'] = W[1]
fil_dict['W_PB2'] = W[2]
self._save(fil_dict, sig.remez(self.N,F,[1, 0, 1], weight = W, Hz = 1,
grid_density = self.grid_density))