def test_plot_frequency_response():
coefs = design_fir_filter(FS, 'bandpass', (8, 12), 3)
f_db, db = compute_frequency_response(coefs, a_vals=1, fs=FS)
plot_frequency_response(f_db, db,
save_fig=True, file_path=TEST_PLOTS_PATH,
file_name='test_plot_frequency_response.png')
def test_plot_frequency_response():
coefs = design_fir_filter(2000, 1000, 'bandpass', (8, 12), 3)
f_db, db = compute_frequency_response(coefs, a_vals=1, fs=1000)
plot_frequency_response(f_db, db)
assert True
def test_plot_filter_properties():
fs = 1000
coefs = design_fir_filter(2000, fs, 'bandpass', (8, 12), 3)
f_db, db = compute_frequency_response(coefs, a_vals=1, fs=1000)
plot_filter_properties(f_db, db, fs, coefs)
assert True
def filter_signal_iir(sig,
fs,
pass_type,
f_range,
butterworth_order,
print_transitions=False,
plot_properties=False,
return_filter=False):
"""Apply an IIR filter to a signal.
Parameters
----------
sig : array
Time series to be filtered.
fs : float
Sampling rate, in Hz.
pass_type : {'bandpass', 'bandstop', 'lowpass', 'highpass'}
Which kind of filter to apply:
* 'bandpass': apply a bandpass filter
* 'bandstop': apply a bandstop (notch) filter
* 'lowpass': apply a lowpass filter
* 'highpass' : apply a highpass filter
f_range : tuple of (float, float) or float
Cutoff frequency(ies) used for filter, specified as f_lo & f_hi.
For 'bandpass' & 'bandstop', must be a tuple.
For 'lowpass' or 'highpass', can be a float that specifies pass frequency, or can be
a tuple and is assumed to be (None, f_hi) for 'lowpass', and (f_lo, None) for 'highpass'.
butterworth_order : int
Order of the butterworth filter, if using an IIR filter.
See input 'N' in scipy.signal.butter.
print_transitions : bool, optional, default: False
If True, print out the transition and pass bandwidths.
plot_properties : bool, optional, default: False
If True, plot the properties of the filter, including frequency response and/or kernel.
return_filter : bool, optional, default: False
If True, return the filter coefficients of the IIR filter.
Returns
-------
sig_filt : 1d array
Filtered time series.
filter_coefs : tuple of (1d array, 1d array)
Filter coefficients of the IIR filter, as (b_vals, a_vals).
Only returned if `return_filter` is True.
"""
# Design filter
b_vals, a_vals = design_iir_filter(fs, pass_type, f_range,
butterworth_order)
# Check filter properties: compute transition bandwidth & run checks
check_filter_properties(b_vals,
a_vals,
fs,
pass_type,
f_range,
verbose=print_transitions)
# Remove any NaN on the edges of 'sig'
sig, sig_nans = remove_nans(sig)
# Apply filter
sig_filt = apply_iir_filter(sig, b_vals, a_vals)
# Add NaN back on the edges of 'sig', if there were any at the beginning
sig_filt = restore_nans(sig_filt, sig_nans)
# Plot frequency response, if desired
if plot_properties:
f_db, db = compute_frequency_response(b_vals, a_vals, fs)
plot_frequency_response(f_db, db)
if return_filter:
return sig_filt, (b_vals, a_vals)
else:
return sig_filt
def check_filter_properties(b_vals, a_vals, fs, pass_type, f_range, transitions=(-20, -3), verbose=True):
"""Check a filters properties, including pass band and transition band.
Parameters
----------
b_vals : 1d array
B value filter coefficients for a filter.
a_vals : 1d array
A value filter coefficients for a filter.
fs : float
Sampling rate, in Hz.
pass_type : {'bandpass', 'bandstop', 'lowpass', 'highpass'}
Which kind of filter to apply:
* 'bandpass': apply a bandpass filter
* 'bandstop': apply a bandstop (notch) filter
* 'lowpass': apply a lowpass filter
* 'highpass' : apply a highpass filter
f_range : tuple of (float, float) or float
Cutoff frequency(ies) used for filter, specified as f_lo & f_hi.
For 'bandpass' & 'bandstop', must be a tuple.
For 'lowpass' or 'highpass', can be a float that specifies pass frequency, or can be
a tuple and is assumed to be (None, f_hi) for 'lowpass', and (f_lo, None) for 'highpass'.
transitions : tuple of (float, float), optional, default: (-20, -3)
Cutoffs, in dB, that define the transition band.
verbose : bool, optional, default: True
Whether to print out transition and pass bands.
Returns
-------
passes : bool
Whether all the checks pass. False if one or more checks fail.
"""
# Import utility functions inside function to avoid circular imports
from neurodsp.filt.utils import (compute_frequency_response,
compute_pass_band, compute_transition_band)
# Initialize variable to keep track if all checks pass
passes = True
# Compute the frequency response
f_db, db = compute_frequency_response(b_vals, a_vals, fs)
# Check that frequency response goes below transition level (has significant attenuation)
if np.min(db) >= transitions[0]:
passes = False
warn('The filter attenuation never goes below {} dB.'\
'Increase filter length.'.format(transitions[0]))
# If there is no attenuation, cannot calculate bands, so return here
return passes
# Check that both sides of a bandpass have significant attenuation
if pass_type == 'bandpass':
if db[0] >= transitions[0] or db[-1] >= transitions[0]:
passes = False
warn('The low or high frequency stopband never gets attenuated by'\
'more than {} dB. Increase filter length.'.format(abs(transitions[0])))
# Compute pass & transition bandwidth
pass_bw = compute_pass_band(fs, pass_type, f_range)
transition_bw = compute_transition_band(f_db, db, transitions[0], transitions[1])
# Raise warning if transition bandwidth is too high
if transition_bw > pass_bw:
passes = False
warn('Transition bandwidth is {:.1f} Hz. This is greater than the desired'\
'pass/stop bandwidth of {:.1f} Hz'.format(transition_bw, pass_bw))
# Print out transition bandwidth and pass bandwidth to the user
if verbose:
print('Transition bandwidth is {:.1f} Hz.'.format(transition_bw))
print('Pass/stop bandwidth is {:.1f} Hz.'.format(pass_bw))
return passes
def test_plot_frequency_response():
coefs = design_fir_filter(FS, 'bandpass', (8, 12), 3)
f_db, db = compute_frequency_response(coefs, a_vals=1, fs=FS)
plot_frequency_response(f_db, db)
def filter_signal_fir(sig, fs, pass_type, f_range, n_cycles=3, n_seconds=None, remove_edges=True,
print_transitions=False, plot_properties=False, return_filter=False):
"""Apply an FIR filter to a signal.
Parameters
----------
sig : array
Time series to be filtered.
fs : float
Sampling rate, in Hz.
pass_type : {'bandpass', 'bandstop', 'lowpass', 'highpass'}
Which kind of filter to apply:
* 'bandpass': apply a bandpass filter
* 'bandstop': apply a bandstop (notch) filter
* 'lowpass': apply a lowpass filter
* 'highpass' : apply a highpass filter
f_range : tuple of (float, float) or float
Cutoff frequency(ies) used for filter, specified as f_lo & f_hi.
For 'bandpass' & 'bandstop', must be a tuple.
For 'lowpass' or 'highpass', can be a float that specifies pass frequency, or can be
a tuple and is assumed to be (None, f_hi) for 'lowpass', and (f_lo, None) for 'highpass'.
n_cycles : float, optional, default: 3
Length of filter, in number of cycles, defined at the 'f_lo' frequency.
This parameter is overwritten by `n_seconds`, if provided.
n_seconds : float, optional
Length of filter, in seconds. This parameter overwrites `n_cycles`.
remove_edges : bool, optional
If True, replace samples within half the kernel length to be np.nan.
print_transitions : bool, optional, default: False
If True, print out the transition and pass bandwidths.
plot_properties : bool, optional, default: False
If True, plot the properties of the filter, including frequency response and/or kernel.
return_filter : bool, optional, default: False
If True, return the filter coefficients of the FIR filter.
Returns
-------
sig_filt : array
Filtered time series.
filter_coefs : 1d array
Filter coefficients of the FIR filter. Only returned if `return_filter` is True.
Examples
--------
Apply a band pass FIR filter to a simulated signal:
>>> from neurodsp.sim import sim_combined
>>> sig = sim_combined(n_seconds=10, fs=500,
... components={'sim_powerlaw': {}, 'sim_oscillation' : {'freq': 10}})
>>> filt_sig = filter_signal_fir(sig, fs=500, pass_type='bandpass', f_range=(1, 25))
Apply a high pass FIR filter to a signal, with a specified number of cycles:
>>> sig = sim_combined(n_seconds=10, fs=500,
... components={'sim_powerlaw': {}, 'sim_oscillation' : {'freq': 10}})
>>> filt_sig = filter_signal_fir(sig, fs=500, pass_type='highpass', f_range=(2, None), n_cycles=5)
"""
# Design filter & check that the length is okay with signal
filter_coefs = design_fir_filter(fs, pass_type, f_range, n_cycles, n_seconds)
check_filter_length(sig.shape[-1], len(filter_coefs))
# Check filter properties: compute transition bandwidth & run checks
check_filter_properties(filter_coefs, 1, fs, pass_type, f_range, verbose=print_transitions)
# Remove any NaN on the edges of 'sig'
sig, sig_nans = remove_nans(sig)
# Apply filter
sig_filt = apply_fir_filter(sig, filter_coefs)
# Remove edge artifacts
if remove_edges:
sig_filt = remove_filter_edges(sig_filt, len(filter_coefs))
# Add NaN back on the edges of 'sig', if there were any at the beginning
sig_filt = restore_nans(sig_filt, sig_nans)
# Plot filter properties, if specified
if plot_properties:
f_db, db = compute_frequency_response(filter_coefs, 1, fs)
plot_filter_properties(f_db, db, fs, filter_coefs)
if return_filter:
return sig_filt, filter_coefs
else:
return sig_filt
def filter_signal_iir(sig,
fs,
pass_type,
f_range,
butterworth_order,
print_transitions=False,
plot_properties=False,
return_filter=False):
"""Apply an IIR filter to a signal.
Parameters
----------
sig : array
Time series to be filtered.
fs : float
Sampling rate, in Hz.
pass_type : {'bandpass', 'bandstop', 'lowpass', 'highpass'}
Which kind of filter to apply:
* 'bandpass': apply a bandpass filter
* 'bandstop': apply a bandstop (notch) filter
* 'lowpass': apply a lowpass filter
* 'highpass' : apply a highpass filter
f_range : tuple of (float, float) or float
Cutoff frequency(ies) used for filter, specified as f_lo & f_hi.
For 'bandpass' & 'bandstop', must be a tuple.
For 'lowpass' or 'highpass', can be a float that specifies pass frequency, or can be
a tuple and is assumed to be (None, f_hi) for 'lowpass', and (f_lo, None) for 'highpass'.
butterworth_order : int
Order of the butterworth filter, if using an IIR filter.
See input 'N' in scipy.signal.butter.
print_transitions : bool, optional, default: False
If True, print out the transition and pass bandwidths.
plot_properties : bool, optional, default: False
If True, plot the properties of the filter, including frequency response and/or kernel.
return_filter : bool, optional, default: False
If True, return the second order series coefficients of the IIR filter.
Returns
-------
sig_filt : 1d array
Filtered time series.
sos : 2d array
Second order series coefficients of the IIR filter. Has shape of (n_sections, 6).
Only returned if `return_filter` is True.
Examples
--------
Apply a bandstop IIR filter to a simulated signal:
>>> from neurodsp.sim import sim_combined
>>> sig = sim_combined(n_seconds=10, fs=500,
... components={'sim_powerlaw': {}, 'sim_oscillation' : {'freq': 10}})
>>> filt_sig = filter_signal_iir(sig, fs=500, pass_type='bandstop',
... f_range=(55, 65), butterworth_order=7)
"""
# Design filter
sos = design_iir_filter(fs, pass_type, f_range, butterworth_order)
# Check filter properties: compute transition bandwidth & run checks
check_filter_properties(sos,
None,
fs,
pass_type,
f_range,
verbose=print_transitions)
# Remove any NaN on the edges of 'sig'
sig, sig_nans = remove_nans(sig)
# Apply filter
sig_filt = apply_iir_filter(sig, sos)
# Add NaN back on the edges of 'sig', if there were any at the beginning
sig_filt = restore_nans(sig_filt, sig_nans)
# Plot frequency response, if desired
if plot_properties:
f_db, db = compute_frequency_response(sos, None, fs)
plot_frequency_response(f_db, db)
if return_filter:
return sig_filt, sos
else:
return sig_filt
butterworth_order = 12 # Design the filter, getting the second-order series (sos) values for the filter sos = design_iir_filter(fs, pass_type, f_range, butterworth_order) ################################################################################################### # # Now that we have our filter definition, we can evaluate our filter. # # Next, we can calculate the frequency response, :math:`b_k`, for our IIR filter. # ################################################################################################### # Compute the frequency response for the IIR filter f_db, db = compute_frequency_response(sos, None, fs) # Plot the frequency response plot_frequency_response(f_db, db) ################################################################################################### # # Above, we can see our frequency response of our filter, which shows us how different # frequencies are affected by our filter. Ideally, we want zero attenuation in our passband, # and a lot of attenuation in the stopband(s). # # Another way to quantify these properties is the transition band, which is bandwidth (in Hz) # that it takes for the filter to change from high to low attenuation. This quantifies how # sharp the transition is between stopband and passband. By default, transition bands are computed # as the range between -20 dB and -3 dB attenuation, but you can also customize these values. #
n_cycles = 3 # Design the filter coefficients for a specified filter filter_coefs = design_fir_filter(fs, pass_type, f_range, n_cycles=n_cycles) ################################################################################################### # # Now that we have our filter coefficients, we can evaluate our filter. # # Next, we can calculate the frequency response, :math:`b_k`, for our alpha bandpass filter. # ################################################################################################### # Compute the frequency response of the filter f_db, db = compute_frequency_response(filter_coefs, 1, fs) # Plot the filter properties plot_filter_properties(f_db, db, fs, filter_coefs) ################################################################################################### # # On the right is the impulse response, or the filter kernel. This is a visualization of our # filter coefficients, which also demonstrates the activity of our filter for a single point # (the impulse response). # # On the left, we can see the frequency response of our filter, which shows us how different # frequencies are affected by our filter. Ideally, we want zero attenuation in our passband, # and a lot of attenuation in the stopband(s). # # Another way to quantify these properties is the transition band, which is bandwidth (in Hz)
