aspen.processings.modulation_power_spectrum¶

Calculate a modulation power spectrum

Functions

modulation_power_spectrum(x[, …])

Modulation Power Spectrum.

Classes

ModulationPowerSpectrum([…])

Modulation Power Spectrum.

class aspen.processings.modulation_power_spectrum.ModulationPowerSpectrum(modulation_power_spectrum_spec_samp_freq=1000, modulation_power_spectrum_gauss_window_alpha=3, modulation_power_spectrum_spacing_freq=50, modulation_power_spectrum_lower_freq=0, modulation_power_spectrum_upper_freq=8000, modulation_power_spectrum_spec_normalize=True, modulation_power_spectrum_spec_db_range=50, modulation_power_spectrum_fft2_win_duration=100, modulation_power_spectrum_fft2_win_shift=0, modulation_power_spectrum_backend='librosa', samp_freq=16000)[source]¶

Bases: aspen.interfaces.abs_common_interface.AbsCommonInterface, aspen.interfaces.abs_processing_interface.AbsProcessingInterface

Modulation Power Spectrum.

This class is heavily inspired by soundsig (https://github.com/theunissenlab/soundsig).

Parameters

modulation_power_spectrum_spec_samp_freq¶ (int) – Sampling frequency in spectrogram space. Defaults to 1000.
modulation_power_spectrum_gauss_window_alpha¶ (float) – The parameter to generate Gaussian window. The detail is shown in MATLAB gaussian window method (https://www.mathworks.com/help/signal/ref/gausswin.html). Defaults to 3.
modulation_power_spectrum_spacing_freq¶ (float) – The time-frequency scale for the spectrogram in Hz. This variable determines the width of the gaussian window to calculate the SFTF. Defaults to 50.
modulation_power_spectrum_lower_freq¶ (float) – Lower frequency in the spectrogram to save space. Defaults to 0.
modulation_power_spectrum_upper_freq¶ (float) – Upper frequency in the spectrogram to save space. Upper limit is the half of sampling_frequency. Defaults to 8000.
modulation_power_spectrum_spec_normalize¶ (bool) – The flag of normalizing spectrogram resulted from STFT. Defaults to True.
modulation_power_spectrum_spec_db_range¶ (float) – The range to narrow down the spectrogram amplitude for making it easier to visualize. Defaults to 50.
modulation_power_spectrum_fft2_win_duration¶ (float) – The duration of gaussian window for 2D-FFT in millisecond. If the value is 0, fft2 is executed withoud window-shifting. Defaults to 100.
modulation_power_spectrum_fft2_win_shift¶ (int) – The number of points to shift segments for 2D-FFT. Defaults is 0 that means (wduration - 1) // 6 (wduration is fft2_win_duration in sample).
modulation_power_spectrum_backend¶ (str) – The library to calculate STFT. The choices are “librosa” or “scipy”. Defaults to “librosa”.
samp_freq¶ (int) – Sampling frequency. Defaults to 16000.

__call__(x)[source]¶

Calculate modulation power spectrum

Parameters: x¶ (ndarray) – Input signal
Return type: ndarray
Returns: Return the modulation power spectrum.

static add_arguments(parser)[source]¶: add arguments

classmethod load_class_kwargs(args)¶

Return the kwargs dict for class __init__ from parsed arguments

Parameters: args¶ (Namespace) – (config)argparse arguments
Return type: dict
Returns: kwargs for class __init__

raw_mps()[source]¶

Return the listed result of 2-D discrete Fourier transform.

Return type: List
Returns: the listed result of 2-D discrete Fourier transform.

spectral_modulation_freq()[source]¶

Return the modulation power spectrum sample spectral modulation frequency.

Return type: ndarray
Returns: the modulation power spectrum sample spectral modulation frequency.

stft_parameters()[source]¶

Return the pameters for short-time Fourier transform.

Return type: Dict
Returns: the pameters for short-time Fourier transform.

temporal_modulation_freq()[source]¶

Return the modulation power spectrum sample temporal modulation frequency.

Return type: ndarray
Returns: the modulation power spectrum sample temporal modulation frequency.

aspen.processings.modulation_power_spectrum.modulation_power_spectrum(x, spec_samp_freq=1000, gauss_window_alpha=3, spacing_freq=50, lower_freq=0, upper_freq=8000, spec_normalize=True, spec_db_range=50, fft2_win_duration=100, fft2_win_shift=0, backend='librosa', samp_freq=16000)[source]¶

Modulation Power Spectrum.

This method is heavily inspired by soundsig (https://github.com/theunissenlab/soundsig).

Parameters

x¶ (ndarray) – Input signal
spec_samp_freq¶ (int) – Sampling frequency in spectrogram space. Defaults to 1000.
gauss_window_alpha¶ (float) – The parameter to generate Gaussian window. The detail is shown in MATLAB gaussian window method (https://www.mathworks.com/help/signal/ref/gausswin.html). Defaults to 3.
spacing_freq¶ (float) – The time-frequency scale for the spectrogram in Hz. This variable determines the width of the gaussian window to calculate the SFTF. Defaults to 50.
lower_freq¶ (float) – Lower frequency in the spectrogram to save space. Defaults to 0.
upper_freq¶ (float) – Upper frequency in the spectrogram to save space. Upper limit is the half of sampling_frequency. Defaults to 8000.
spec_normalize¶ (bool) – The flag of normalizing spectrogram resulted from STFT. Defaults to True.
spec_db_range¶ (float) – The range to narrow down the spectrogram amplitude for making it easier to visualize. Defaults to 50.
fft2_win_duration¶ (float) – The duration of gaussian window for 2D-FFT in millisecond. If the value is 0, fft2 is executed withoud window-shifting. Defaults to 100.
fft2_win_shift¶ (int) – The number of points to shift segments for 2D-FFT. Defaults is 0 that means (wduration - 1) // 6 (wduration is fft2_win_duration in sample).
backend¶ (str) – The library to calculate STFT. The choices are “librosa” or “scipy”. Defaults to “librosa”.
samp_freq¶ (int) – Sampling frequency. Defaults to 16000.

Return type

Tuple[Dict, ndarray, ndarray, List, ndarray]

Returns

Return the pameters for short-time Fourier transform, the modulation power spectrum sample spectral modulation frequency, the modulation power spectrum sample temporal modulation frequency, the listed result of 2-D discrete Fourier transform and the modulation power spectrum.