.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "auto_examples/stride_segmentation/roth_hmm_stride_segmentation.py"
.. LINE NUMBERS ARE GIVEN BELOW.

.. only:: html

    .. note::
        :class: sphx-glr-download-link-note

        Click :ref:`here <sphx_glr_download_auto_examples_stride_segmentation_roth_hmm_stride_segmentation.py>`
        to download the full example code

.. rst-class:: sphx-glr-example-title

.. _sphx_glr_auto_examples_stride_segmentation_roth_hmm_stride_segmentation.py:


.. _example_roth_stride_segmentation:

HMM stride segmentation - Prediction with pre-trained model
===========================================================

This example illustrates how a Hidden Markov Model (HMM) implemented by the
:class:`~gaitmap.stride_segmentation.hmm.HmmStrideSegmentation` can be used to detect strides in a continuous signal of
an IMU signal.
The used implementation is based on the work of Roth et al [1]_

.. [1] Roth, N., Küderle, A., Ullrich, M., Gladow, T., Marxreiter F., Klucken, J., Eskofier, B. & Kluge F. (2021).
   Hidden Markov Model based Stride Segmentation on Unsupervised Free-living Gait Data in Parkinson’s Disease Patients.
   Journal of NeuroEngineering and Rehabilitation, (JNER).

.. GENERATED FROM PYTHON SOURCE LINES 16-22

.. code-block:: default


    import matplotlib.pyplot as plt
    import numpy as np

    np.random.seed(0)








.. GENERATED FROM PYTHON SOURCE LINES 23-29

Getting some example data
--------------------------

For this we take some example data that contains the regular walking movement during a 2x20m walk test of a healthy
subject. The IMU signals are already rotated so that they align with the gaitmap SF coordinate system.
The data contains information from two sensors - one from the right and one from the left foot.

.. GENERATED FROM PYTHON SOURCE LINES 29-35

.. code-block:: default

    from gaitmap.example_data import get_healthy_example_imu_data

    data = get_healthy_example_imu_data()
    sampling_rate_hz = 204.8
    data.sort_index(axis=1).head(1)






.. raw:: html

    <div class="output_subarea output_html rendered_html output_result">
    <div>
    <style scoped>
        .dataframe tbody tr th:only-of-type {
            vertical-align: middle;
        }

        .dataframe tbody tr th {
            vertical-align: top;
        }

        .dataframe thead tr th {
            text-align: left;
        }
    </style>
    <table border="1" class="dataframe">
      <thead>
        <tr>
          <th>sensor</th>
          <th colspan="6" halign="left">left_sensor</th>
          <th colspan="6" halign="left">right_sensor</th>
        </tr>
        <tr>
          <th>axis</th>
          <th>acc_x</th>
          <th>acc_y</th>
          <th>acc_z</th>
          <th>gyr_x</th>
          <th>gyr_y</th>
          <th>gyr_z</th>
          <th>acc_x</th>
          <th>acc_y</th>
          <th>acc_z</th>
          <th>gyr_x</th>
          <th>gyr_y</th>
          <th>gyr_z</th>
        </tr>
      </thead>
      <tbody>
        <tr>
          <th>0.0</th>
          <td>0.880811</td>
          <td>2.762208</td>
          <td>9.40865</td>
          <td>-0.112402</td>
          <td>-0.032157</td>
          <td>-0.062261</td>
          <td>0.311553</td>
          <td>-2.398646</td>
          <td>9.513275</td>
          <td>-0.323037</td>
          <td>0.084604</td>
          <td>-0.025288</td>
        </tr>
      </tbody>
    </table>
    </div>
    </div>
    <br />
    <br />

.. GENERATED FROM PYTHON SOURCE LINES 36-42

Preparing the data
------------------
The HMM only makes use of the gyro information.
Further, if you use this model, your data is expected to be in the gaitmap body-frame to be able to use the
same model for the left and the right foot.
Therefore, we need to transform the dataset into the body frame.

.. GENERATED FROM PYTHON SOURCE LINES 42-47

.. code-block:: default

    from gaitmap.utils.coordinate_conversion import convert_to_fbf

    # We use the `..._like` parameters to identify the data of the left and the right foot based on the name of the sensor.
    bf_data = convert_to_fbf(data, left_like="left_", right_like="right_")








.. GENERATED FROM PYTHON SOURCE LINES 48-53

Selecting a pre-trained model
-----------------------------
This library ships with pre-trained models that can be directly used for prediction/ segmentation.
It is generated based on manually segmented strides from healthy participants and PD patients.
We can load the model a look at some of its parameters

.. GENERATED FROM PYTHON SOURCE LINES 53-63

.. code-block:: default


    from gaitmap.stride_segmentation.hmm import PreTrainedRothSegmentationModel

    roth_hmm_model = PreTrainedRothSegmentationModel()

    print(f"Number of states, stride-model: {roth_hmm_model.stride_model.n_states:d}")
    print(f"Number of states, transition-model: {roth_hmm_model.transition_model.n_states:d}")
    np.set_printoptions(precision=3, linewidth=180, suppress=True)
    print(f"Transition matrix:\n{roth_hmm_model.model.dense_transition_matrix()[0:-2, 0:-2]}")





.. rst-class:: sphx-glr-script-out

 .. code-block:: none

    Number of states, stride-model: 20
    Number of states, transition-model: 5
    Transition matrix:
    [[0.882 0.036 0.    0.054 0.    0.028 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.975 0.025 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.975 0.025 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.942 0.058 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.085 0.    0.    0.    0.915 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.589 0.411 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.667 0.333 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.68  0.32  0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.694 0.306 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.696 0.304 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.68  0.32  0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.691 0.309 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.682 0.318 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.653 0.347 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.674 0.326 0.    0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.69  0.31  0.    0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.668 0.332 0.    0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.764 0.236 0.    0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.849 0.151 0.    0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.765 0.235 0.    0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.632 0.368 0.    0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.594 0.406 0.    0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.564 0.436 0.   ]
     [0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.398 0.602]
     [0.039 0.    0.    0.    0.    0.369 0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.    0.592]]




.. GENERATED FROM PYTHON SOURCE LINES 64-68

Predicting hidden states / Stride borders
-----------------------------------------
To use this model to actually segment the data, we wrap it in the `HmmStrideSegmentation` class.
This class provides a interface and post-processing similar to other Stride Segmentation algorithms.

.. GENERATED FROM PYTHON SOURCE LINES 68-73

.. code-block:: default

    from gaitmap.stride_segmentation.hmm import HmmStrideSegmentation

    hmm_seg = HmmStrideSegmentation(roth_hmm_model, snap_to_min_win_ms=300, snap_to_min_axis="gyr_ml")
    hmm_seg = hmm_seg.segment(bf_data, sampling_rate_hz=sampling_rate_hz)








.. GENERATED FROM PYTHON SOURCE LINES 74-78

Inspecting the results
----------------------
The main output is the `stride_list_`, which contains the start and the end of all identified strides.
As we passed a dataset with two sensors, the output will be a dictionary.

.. GENERATED FROM PYTHON SOURCE LINES 78-82

.. code-block:: default

    stride_list_left = hmm_seg.stride_list_["left_sensor"]
    print(f"{len(stride_list_left)} strides were detected.")
    stride_list_left.head()





.. rst-class:: sphx-glr-script-out

 .. code-block:: none

    29 strides were detected.


.. raw:: html

    <div class="output_subarea output_html rendered_html output_result">
    <div>
    <style scoped>
        .dataframe tbody tr th:only-of-type {
            vertical-align: middle;
        }

        .dataframe tbody tr th {
            vertical-align: top;
        }

        .dataframe thead th {
            text-align: right;
        }
    </style>
    <table border="1" class="dataframe">
      <thead>
        <tr style="text-align: right;">
          <th></th>
          <th>start</th>
          <th>end</th>
        </tr>
        <tr>
          <th>s_id</th>
          <th></th>
          <th></th>
        </tr>
      </thead>
      <tbody>
        <tr>
          <th>0</th>
          <td>364</td>
          <td>584</td>
        </tr>
        <tr>
          <th>1</th>
          <td>584</td>
          <td>802</td>
        </tr>
        <tr>
          <th>2</th>
          <td>802</td>
          <td>1023</td>
        </tr>
        <tr>
          <th>3</th>
          <td>1023</td>
          <td>1242</td>
        </tr>
        <tr>
          <th>4</th>
          <td>1242</td>
          <td>1458</td>
        </tr>
      </tbody>
    </table>
    </div>
    </div>
    <br />
    <br />

.. GENERATED FROM PYTHON SOURCE LINES 83-96

To get a better understanding of the results, we can plot additional information about the results.
The top row shows the `gyr_ml` axis with the segmented strides plotted on top.
They are postprocessed to snap to the closed data minimum.
In the second row the predicted hidden state sequence of the HMM is plotted (this is the transformed version, matching
the input signal).
Each transition from the last (n=25) to the first (n=5) stride state marks a potential start/end of a stride.
The second plot shows the results in the feature space (which will depend on the feature space setting during the
training step).
Here this is a downsampled and filtered representative of the gyr_ml signal as well as its window based gradient.
All features are z-transformed (note we z-transform the new data independently from the training data).
Again, the predicted hidden state sequence is plotted together with the data.

Only the first couple of strides of the left foot are shown.

.. GENERATED FROM PYTHON SOURCE LINES 96-139

.. code-block:: default


    sensor = "left_sensor"

    fig, axs = plt.subplots(nrows=2, sharex=True, figsize=(10, 5))
    axs[0].set_title("gaitmap Body Frame Dataset")
    axs[0].plot(bf_data.reset_index(drop=True)[sensor]["gyr_ml"])
    for start, end in hmm_seg.stride_list_["left_sensor"].to_numpy():
        axs[0].axvline(start, c="r")
        axs[0].axvline(end, c="r")
        axs[0].axvspan(start, end, alpha=0.2)
    axs[0].set_ylabel("gyr-ml [deg/s]")

    axs[1].set_title("Predicted Hidden State Sequence")
    axs[1].plot(hmm_seg.hidden_state_sequence_[sensor])
    for start, end in hmm_seg.matches_start_end_original_[sensor]:
        axs[1].axvline(start, c="g")
        axs[1].axvline(end, c="g")
        axs[1].axvspan(start, end, alpha=0.2)
    axs[1].set_ylabel("Hidden State [N]")

    axs[1].set_xlabel("Samples @ %d Hz" % sampling_rate_hz)
    plt.xlim([0, 5000])
    fig.tight_layout()
    plt.show()

    fig, ax1 = plt.subplots(figsize=(10, 3))
    plt.title("HMM Feature Space")
    ax1.set_xlabel(f"Samples Features Space @ {hmm_seg.model.feature_transform.sampling_rate_feature_space_hz} Hz")
    ax1.set_ylabel("Z-Transform [a.u.]")
    feature_space_date = hmm_seg.result_model_[sensor].feature_space_data_
    ax1.plot(feature_space_date)
    ax1.legend(feature_space_date.columns.to_list())

    ax2 = ax1.twinx()
    ax2.set_ylabel("Hidden State Sequence", color="tab:green")
    hidden_state_sequence_feature_space = hmm_seg.result_model_[sensor].hidden_state_sequence_feature_space_
    ax2.plot(hidden_state_sequence_feature_space, color="tab:green")
    ax2.tick_params(axis="y", labelcolor="tab:green")

    plt.xlim([0, 500])
    fig.tight_layout()
    plt.show()




.. rst-class:: sphx-glr-horizontal


    *

      .. image-sg:: /auto_examples/stride_segmentation/images/sphx_glr_roth_hmm_stride_segmentation_001.png
         :alt: gaitmap Body Frame Dataset, Predicted Hidden State Sequence
         :srcset: /auto_examples/stride_segmentation/images/sphx_glr_roth_hmm_stride_segmentation_001.png
         :class: sphx-glr-multi-img

    *

      .. image-sg:: /auto_examples/stride_segmentation/images/sphx_glr_roth_hmm_stride_segmentation_002.png
         :alt: HMM Feature Space
         :srcset: /auto_examples/stride_segmentation/images/sphx_glr_roth_hmm_stride_segmentation_002.png
         :class: sphx-glr-multi-img






.. rst-class:: sphx-glr-timing

   **Total running time of the script:** ( 0 minutes  8.250 seconds)

**Estimated memory usage:**  11 MB


.. _sphx_glr_download_auto_examples_stride_segmentation_roth_hmm_stride_segmentation.py:

.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-example


    .. container:: sphx-glr-download sphx-glr-download-python

      :download:`Download Python source code: roth_hmm_stride_segmentation.py <roth_hmm_stride_segmentation.py>`

    .. container:: sphx-glr-download sphx-glr-download-jupyter

      :download:`Download Jupyter notebook: roth_hmm_stride_segmentation.ipynb <roth_hmm_stride_segmentation.ipynb>`


.. only:: html

 .. rst-class:: sphx-glr-signature

    `Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_