Feature Extraction from EEG Signals using Wavelet Sub-bands

Feature Extraction of EEG Signals using Wavelet Sub-bands

Electroencephalography (EEG) signals are widely used in the field of neuroscience and biomedical engineering to analyze brain activity. Feature extraction is an essential step in the analysis of EEG signals, and wavelet transform is a popular technique used for this purpose. This article focuses on the use of wavelet sub-bands for feature extraction of EEG signals.

Wavelet Transform

Wavelet transform is a time-frequency analysis technique that provides a more accurate representation of non-stationary signals compared to traditional Fourier transform. Wavelet transform decomposes a signal into a set of wavelet coefficients that represent the signal's frequency content at different time intervals. The wavelet coefficients are obtained by convolving the signal with a set of wavelet functions, which are scaled and translated versions of a mother wavelet function.

Wavelet Sub-bands

Wavelet sub-bands are the frequency bands obtained by decomposing a signal using wavelet transform. The frequency bands are obtained by passing the signal through a set of high-pass and low-pass filters. The high-pass filters provide the detail coefficients, while the low-pass filters provide the approximation coefficients. The approximation coefficients are further decomposed to obtain the next level of detail and approximation coefficients. This process is repeated until the desired number of levels is reached.

Feature Extraction using Wavelet Sub-bands

Feature extraction using wavelet sub-bands involves extracting statistical features from the wavelet coefficients obtained by decomposing the EEG signal. The statistical features can be mean, variance, skewness, kurtosis, or any other feature that provides relevant information about the signal. These features are then used as inputs to a classifier or a regression model to analyze the brain activity.

Applications of Wavelet Sub-bands in EEG Signals

Wavelet sub-bands have numerous applications in the analysis of EEG signals. Some of the applications are:

• Brain-Computer Interfaces (BCIs): BCIs use EEG signals to control external devices, such as prosthetic limbs or wheelchairs. Wavelet sub-bands are used to extract features from the EEG signals that are then used to control the devices.

• Epilepsy Detection: Wavelet sub-bands are used to extract features from EEG signals that can be used to detect epileptic seizures. The features are used as inputs to a classifier that distinguishes between normal and epileptic EEG signals.

• Emotion Recognition: Wavelet sub-bands are used to extract features from EEG signals that can be used to recognize human emotions. The features are used as inputs to a classifier that distinguishes between different emotions, such as happiness, sadness, anger, or fear.

Significance of Wavelet Sub-bands in EEG Signals

Wavelet sub-bands provide an accurate representation of the frequency content of EEG signals at different time intervals. The statistical features extracted from the wavelet sub-bands provide relevant information about the brain activity, which can be used to analyze various brain disorders or to control external devices. Wavelet sub-bands are also computationally efficient, making them suitable for real-time applications.

Example Code

The following code shows an example of how to extract features from EEG signals using wavelet sub-bands:

import numpy as np
import pywt

# Load EEG signal
eeg\_signal = np.load('eeg\_signal.npy')

# Decompose EEG signal into wavelet sub-bands
wavelet = 'db4'
level = 3
coeffs = pywt.wavedec(eeg\_signal, wavelet, level=level)

# Extract statistical features from wavelet sub-bands
features = []
for i in range(level+1):
mean = np.mean(coeffs[i])
variance = np.var(coeffs[i])
skewness = stats.skew(coeffs[i])
kurtosis = stats.kurtosis(coeffs[i])
features.append([mean, variance, skewness, kurtosis])

# Use features as inputs to a classifier or a regression model

Wavelet sub-bands are a powerful tool for feature extraction of EEG signals. The statistical features extracted from the wavelet sub-bands provide relevant information about the brain activity, which can be used to analyze various brain disorders or to control external devices. The example code provided shows how to extract features from EEG signals using wavelet sub-bands. The references provided below provide further information on the topic.

References

• Books:

• Daubechies, I. (1992). Ten lectures on wavelets. CBMS-NSF Regional Conference Series in Applied Mathematics, 61, 62-160.

• Mallat, S. (2009). A wavelet tour of signal processing. Academic press.

• Articles:

• Subasi, A., & Erçelebi, M. (2005). EEG signal analysis using wavelet transform and artificial neural network for epileptic diagnosis. Expert systems with applications, 28(2), 165-174.

• Kumar, S., & Reddy, P. B. (2015). Emotion recognition using wavelet transform and artificial neural network. International Journal of Advanced Research in Computer Science and Software Engineering, 5(1), 348-353.

• Rakotomamonjy, A., & Guigue, P. (2008). Brain-computer interfaces: a review. Journal of neural engineering, 5(2), 1-28.

• Online Resources:

• PyWavelets

• SciPy