Spectral analysis of the trials

This example demonstrates how to perform spectral analysis on epochs extracted from a specific subject within the moabb.datasets.Cattan2019_PHMD dataset.

# Authors: Pedro Rodrigues <pedro.rodrigues01@gmail.com>
# Modified by: Gregoire Cattan <gcattan@hotmail.fr>
# License: BSD (3-clause)

import warnings

import matplotlib.pyplot as plt
import numpy as np

from moabb.datasets import Cattan2019_PHMD
from moabb.paradigms import RestingStateToP300Adapter


warnings.filterwarnings("ignore")

Initialization

1. Specify the channel and subject to compute the power spectrum.

2. Create an instance of the moabb.datasets.Cattan2019_PHMD dataset.

3. Create an instance of the moabb.paradigms.RestingStateToP300Adapter paradigm. By default, the data is filtered between 1-35 Hz, and epochs are extracted from 10 to 50 seconds after event tagging.

# Select channel and subject for the remaining of the example.
channel = "Cz"
subject = 1

dataset = Cattan2019_PHMD()
events = ["on", "off"]
paradigm = RestingStateToP300Adapter(events=events, channels=[channel])

Estimate Power Spectral Density

1. Obtain the epochs for the specified subject.

2. Use Welch’s method to estimate the power spectral density.

f, S, _, y = paradigm.psd(subject, dataset)

This is nemar-py 0.3.0.
Preparing to download nm000341 from https://data.nemar.org/
Retrieving 5 of 119 manifest files (16 concurrent downloads).
primary backend failed; falling back to next layer: 'README.md' is not annexed (no sha256/md5 checksum on the manifest entry); the S3 backend has nothing to fetch.

Overall:   0%|          | 0.00/10.6k [00:00<?, ?B/s]
Overall:   6%|▌         | 616/10.6k [00:00<00:01, 5.19kB/s]
Overall:  43%|████▎     | 4.63k/10.6k [00:00<00:00, 34.0kB/s]
Finished downloading nm000341 v1.0.5.
This is nemar-py 0.3.0.
Preparing to download nm000341 from https://data.nemar.org/
Retrieving 5 of 119 manifest files (16 concurrent downloads).

S3:   0%|          | 0.00/23.9M [00:00<?, ?B/s]
S3:   4%|▍         | 1.00M/23.9M [00:00<00:09, 2.45MB/s]
S3:  17%|█▋        | 4.00M/23.9M [00:00<00:02, 9.51MB/s]
S3:  50%|█████     | 12.0M/23.9M [00:00<00:00, 28.2MB/s]
S3:  96%|█████████▋| 23.0M/23.9M [00:00<00:00, 50.5MB/s]
S3: 100%|██████████| 23.9M/23.9M [00:00<00:00, 33.2MB/s]
Finished downloading nm000341 v1.0.5.

Display of the data

Plot the averaged Power Spectral Density (PSD) for each label condition, using the selected channel specified at the beginning of the script.

fig, ax = plt.subplots(facecolor="white", figsize=(8.2, 5.1))
for condition in events:
    mean_power = np.mean(S[y == condition], axis=0).flatten()
    ax.plot(f, 10 * np.log10(mean_power), label=condition)

ax.set_xlim(paradigm.fmin, paradigm.fmax)
ax.set_ylim(100, 135)
ax.set_ylabel("Spectrum Magnitude (dB)", fontsize=14)
ax.set_xlabel("Frequency (Hz)", fontsize=14)
ax.set_title("PSD for Channel " + channel, fontsize=16)
ax.legend()
fig.show()

Estimated memory usage: 608 MB