minerva.transforms.activity_image
=================================

.. py:module:: minerva.transforms.activity_image


Classes
-------

.. autoapisummary::

   minerva.transforms.activity_image.ActivityImageTransforms


Module Contents
---------------

.. py:class:: ActivityImageTransforms(signal_sequences=9, signal_samples=68)

   A class to transform signals into images for Human Activity Recognition (HAR), 
   based on the methodology presented in the paper 
   "Human Activity Recognition using Wearable Sensors by Deep Convolutional Neural Networks."

   Initializes the ActivityImageTransforms class with the given number of signal sequences and samples.

   Parameters:
   -----------
   signal_sequences: int, optional
       The number of signal sequences. Defaults to 9
   signal_samples: int, optional
       The number of samples per signal sequence. Defaults to 68.


   .. py:method:: __call__(x)

      Transforms the input signals into activity images suitable for CNN processing.
          
      This method performs the following steps:
      1. Resamples the input signals to ensure consistent sample length.
      2. Constructs a 2D signal image from the resampled signals.
      3. Applies a 2D Fast Fourier Transform (FFT) to convert the signal image to the frequency domain.
      4. Shifts the zero-frequency component to the center of the spectrum.
      5. Computes the magnitude of the frequency domain representation to form the final activity image.

      Parameters:
      -----------
      x: (List[np.ndarray])
          The raw signal sequences from wearable sensors.

      Returns:
      --------
          tuple: A tuple containing the original signals and the corresponding activity images.



   .. py:method:: get_signal_image(raw_signals)

      Constructs a signal image from the raw signals by arranging them in a specific order.

      This method follows the approach of combining sensor signals to form a 2D image, which is then 
      suitable for CNN processing. The image is constructed by iteratively selecting signal sequences 
      in a way that maximizes the spatial correlation between different sequences.

      Parameters:
      -----------
      raw_signals: np.ndarray
          The raw signal sequences from wearable sensors.

      Returns:
      --------
      signal_image: np.ndarray
          The constructed signal image, ready for further processing by a CNN.



   .. py:method:: resample_signals(x)

      Resamples the signals to a fixed number of samples.

      This resampling step ensures that all signal sequences have the same length, which is necessary 
      for constructing a consistent 2D image for CNN input.

      Parameters:
      -----------
      x: np.array
          The raw signal sequences from wearable sensors.

      Returns:
      --------
      resample_signal: List[np.ndarray]
          The resampled signals, all having the same number of samples.



   .. py:attribute:: signal_samples
      :value: 68



   .. py:attribute:: signal_sequences
      :value: 9