impplemented amplitude check before thresholding inst freq

This commit is contained in:
weygoldt 2023-04-07 13:56:15 +02:00
parent e1e1e75cd2
commit 21c74b3cb0
No known key found for this signature in database
3 changed files with 48 additions and 37 deletions

View File

@ -535,7 +535,7 @@ def mask_low_amplitudes(envelope, threshold):
"""
mask = np.ones_like(envelope, dtype=bool)
mask[envelope < threshold] = np.nan
mask[envelope < threshold] = False
return mask
@ -835,6 +835,14 @@ def chirpdetection(datapath: str, plot: str, debug: str = "false") -> None:
cutoff_frequency=config.baseline_envelope_cutoff,
)
# create a mask that removes areas where amplitudes are very
# because the instantaneous frequency is not reliable there
amplitude_mask = mask_low_amplitudes(
baseline_envelope_unfiltered,
config.baseline_min_amplitude
)
# highpass filter baseline envelope to remove slower
# fluctuations e.g. due to motion envelope
@ -845,14 +853,6 @@ def chirpdetection(datapath: str, plot: str, debug: str = "false") -> None:
highf=config.baseline_envelope_bandpass_highf,
)
# create a mask that removes areas where amplitudes are very
# because the instantaneous frequency is not reliable there
amplitude_mask = mask_low_amplitudes(
baseline_envelope,
config.baseline_min_amplitude
)
# invert baseline envelope to find troughs in the baseline
baseline_envelope = -baseline_envelope
@ -876,15 +876,10 @@ def chirpdetection(datapath: str, plot: str, debug: str = "false") -> None:
# chirp. This phenomenon is only observed on chirps on a narrow
# filtered baseline such as the one we are working with.
baseline_frequency = instantaneous_frequency2(baselineband, data.raw_rate)
(
baseline_frequency_time,
baseline_frequency,
) = instantaneous_frequency(
signal=baselineband,
samplerate=data.raw_rate,
smoothing_window=config.baseline_frequency_smoothing,
baseline_frequency = instantaneous_frequency(
baselineband,
data.raw_rate,
config.baseline_frequency_smoothing
)
# Take the absolute of the instantaneous frequency to invert
@ -902,9 +897,7 @@ def chirpdetection(datapath: str, plot: str, debug: str = "false") -> None:
# to enter normalization, where small changes due to noise
# would be amplified
# if not has_chirp(baseline_frequency[amplitude_mask], config.baseline_frequency_peakheight):
# continue
if not has_chirp(baseline_frequency, config.baseline_frequency_peakheight):
if not has_chirp(baseline_frequency_filtered[amplitude_mask], config.baseline_frequency_peakheight):
continue
# CUT OFF OVERLAP ---------------------------------------------
@ -925,20 +918,24 @@ def chirpdetection(datapath: str, plot: str, debug: str = "false") -> None:
search_envelope_unfiltered = search_envelope_unfiltered[no_edges]
search_envelope = search_envelope[no_edges]
# get instantaneous frequency withoup edges
no_edges_t0 = int(window_edge) / data.raw_rate
no_edges_t1 = baseline_frequency_time[-1] - (
int(window_edge) / data.raw_rate
)
no_edges = (baseline_frequency_time >= no_edges_t0) & (
baseline_frequency_time <= no_edges_t1
)
baseline_frequency_filtered = baseline_frequency_filtered[no_edges]
baseline_frequency = baseline_frequency[no_edges]
baseline_frequency_time = (
baseline_frequency_time[no_edges] + window_start_seconds
)
baseline_frequency_filtered = baseline_frequency_filtered[no_edges]
baseline_frequency_time = current_raw_time
# # get instantaneous frequency withoup edges
# no_edges_t0 = int(window_edge) / data.raw_rate
# no_edges_t1 = baseline_frequency_time[-1] - (
# int(window_edge) / data.raw_rate
# )
# no_edges = (baseline_frequency_time >= no_edges_t0) & (
# baseline_frequency_time <= no_edges_t1
# )
# baseline_frequency_filtered = baseline_frequency_filtered[no_edges]
# baseline_frequency = baseline_frequency[no_edges]
# baseline_frequency_time = (
# baseline_frequency_time[no_edges] + window_start_seconds
# )
# NORMALIZE ---------------------------------------------------

View File

@ -27,7 +27,7 @@ baseline_frequency_smoothing: 3 # instantaneous frequency smoothing
# Feature processing parameters -----------------------------------------------
baseline_frequency_peakheight: 5 # the min peak height of the baseline instfreq
baseline_min_amplitude: 0.1 # the minimal value of the baseline envelope
baseline_min_amplitude: 0.0001 # the minimal value of the baseline envelope
baseline_envelope_cutoff: 25 # envelope estimation cutoff
baseline_envelope_bandpass_lowf: 2 # envelope badpass lower cutoff
baseline_envelope_bandpass_highf: 100 # envelope bandbass higher cutoff

View File

@ -22,6 +22,7 @@ def minmaxnorm(data):
"""
return (data - np.min(data)) / (np.max(data) - np.min(data))
def instantaneous_frequency2(signal: np.ndarray, fs: float, interpolation: str = 'linear') -> np.ndarray:
"""
Compute the instantaneous frequency of a periodic signal using zero crossings and resample the frequency using linear
@ -66,7 +67,8 @@ def instantaneous_frequency(
signal: np.ndarray,
samplerate: int,
smoothing_window: int,
) -> tuple[np.ndarray, np.ndarray]:
interpolation: str = 'linear',
) -> np.ndarray:
"""
Compute the instantaneous frequency of a signal that is approximately
sinusoidal and symmetric around 0.
@ -79,6 +81,8 @@ def instantaneous_frequency(
Samplerate of the signal.
smoothing_window : int
Window size for the gaussian filter.
interpolation : str, optional
Interpolation method to use during resampling. Should be either 'linear' or 'cubic'. Default is 'linear'.
Returns
-------
@ -112,7 +116,17 @@ def instantaneous_frequency(
1 / np.diff(true_zero), smoothing_window
)
return instantaneous_frequency_time, instantaneous_frequency
# Resample the frequency using specified interpolation method to match the dimensions of the input array
orig_len = len(signal)
freq = resample(instantaneous_frequency, orig_len)
if interpolation == 'linear':
freq = np.interp(np.arange(0, orig_len), np.arange(0, orig_len), freq)
elif interpolation == 'cubic':
freq = resample(freq, orig_len, window='cubic')
return freq
def purge_duplicates(