mightypy.signal_processing package

Module contents

mightypy.signal_processing

class PSDDenoiser(threshold: int | float | str | None = 'auto-mean')

Bases: object

PSD (Power Spectral Density) Based Denoiser

This method takes the FFT transform of the signal to calculate PSD based on the PSD results and cutoff threshold the signal is filtered and a FFT inverse is applied to regenerate denoised signal.

Parameters:

threshold (Optional[Union[int, float, str]], optional) – threshold to create cutoff mask, but any threshold can be applied, if it is precalculated by any method chosen by the process, by default auto-mean { auto-mean, auto-max }

Examples

>>> import numpy as np
>>> import matplotlib.pyplot as plt
>>> from mightypy.preprocessing import PSDDenoiser
>>> rng = np.random.default_rng()
>>> fs = 10e3
>>> N = 100
>>> amp = 2 * np.sqrt(2)
>>> freq = 1234.0
>>> noise_power = 0.001 * fs / 2
>>> time = np.arange(N) / fs
>>> X = amp * np.sin(2 * np.pi * freq * time)
>>> X += rng.normal(scale=np.sqrt(noise_power), size=time.shape)

>>> denoiser = PSDDenoiser()
>>> cleaned_signal = denoiser.transform(X)
>>> plt.plot(X, label="noisy")
>>> plt.plot(cleaned_signal, label="cleaned")
>>> plt.title(f"Threshold : {denoiser.threshold}")
>>> plt.legend(loc="best")
>>> plt.show()

>>> denoiser = PSDDenoiser(10)
>>> cleaned_signal = denoiser.transform(X)
>>> plt.plot(X, label='noisy')
>>> plt.plot(cleaned_signal, label='cleaned')
>>> plt.title(f"Threshold : {denoiser.threshold}")
>>> plt.legend(loc='best')
>>> plt.show()

property f_hat: ndarray | None

FFT of input signal

property filtered_f_hat: ndarray | None

filtered FFT of input signal

psd(f_hat: ndarray, tau: int) → ndarray

Power Spectral Density

Parameters:

• f_hat (np.ndarray) – Signal in Frequency Domain

• tau (int) – Interval

Returns:

Power spectrum

Return type:

np.ndarray

property threshold: str | float | int

Threshold calculated by the process

In Power spectrum after the half length it takes the aggregation of the values and use that as a threshold to cutoff frequencies that are insignificant.

Returns:

cutoff threshold value

Return type:

Union[str, float, int]

transform(X: ndarray) → ndarray

Apply PSD

Parameters:

X (np.ndarray) – Input matrix, signal in IOT Terms.

Returns:

Denoised Signal

Return type:

np.ndarray