Psd

Demonstration code for computing power spectral density using Psd().

Link to script: feature_extraction/psd.py

import os
import numpy as np
import matplotlib.pyplot as plt

from nnsa import Psd, print_object_summary, SUPPORTED_RESULT_FILE_TYPES, read_result_from_file, assert_equal, \
    TimeSeries

plt.close('all')

Parameters.

# Print the default parameters of Psd():
print(Psd().default_parameters())

# Descriptions of the parameters are documented in the default_parameters() code.

# Create an instance of the Psd class with custom parameters, overruling some defaults:
psd = Psd(window_length=2,
          df_max=0.1,
          detrend='constant')

# See if the custom parameters were accepted:
print('\nCustom parameters:')
print(psd.parameters)

Main method: psd.

# Now that we have initialized a Psd object with certain parameters, we can run the psd method, which computes the psd:
# Create a random signal.
signal = np.random.rand(10000)
fs = 250
result = psd.psd(signal, fs=fs, unit='uV', axis=-1, verbose=0)  # Note that unit, axis and verbose are optional args.

# The returned object is a PsdResult object, which is a high-level interface for manipulating/visualizing/saving the
# result:
print('\nSummary of result object:')
print_object_summary(result)

PsdResult attributes.

# We could check whether the df does not exceed the df_max parameter that we specified:
print('df_max: {}\ndf: {}'.format(psd.parameters['df_max'], result.df))

# The main attribute is power_density, which contains the actual psd:
print('power_density: {}'.format(result.power_density))

# The frequencies corresponding to power_density can be constructed from df. This is done by the frequencies property:
print('frequencies: {}'.format(result.frequencies))

PsdResult methods.

# Plot the power spectral density.
result.plot(label='result')

# Compute the power in a specific frequency band, with band limits f_low and f_high:
f_low = 0
f_high = result.frequencies[round(len(result.power_density)/2)]
power = result.compute_power(f_low, f_high)
print('Power {:.2f}-{:.2f} Hz = {:.2f} {}'.format(f_low, f_high, power, result.unit.replace('/Hz', '')))

Save the result.

# We can save the result to any supported file. To list the supported file types/extensions:
print('Supported result file types: {}'.format(SUPPORTED_RESULT_FILE_TYPES))

# E.g. save as hdf5 (the file type to save is automatically detected from the extension of the filename).
filename = 'temp_result.hdf5'
result.save_to_file(filename)

# We can reload the result back to a PsdResult object:
result_loaded = read_result_from_file(filename)

# Remove temporary file.
os.remove(filename)

# Verify that the loaded object is equal to the original object:
assert_equal(actual=result_loaded, desired=result)  # Will raise an AssertionError if not equal.
print('Loaded result object is equal to original object.')

Psd of TimeSeries.

# We can also directly call psd on a TimeSeries object, where we specify the parameters to the psd function as kwargs:
ts = TimeSeries(signal=signal, fs=fs, label='Cz', unit='uV', info={'source': 'random'})
result_ts = ts.psd(window_length=2,
                   df_max=0.1,
                   detrend='constant')

# If we chose the same parameters, result and result_ts should be equal, except for the data_info object of the result,
# which contains the info dictionary of the TimeSeries:
print('result:\n', result)
print('result_ts:\n', result_ts)

result_ts.plot('--', label='result_ts')
plt.legend()