Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
ICA (Independent Component Analysis) for EEG signal analysis
Transcript of ICA (Independent Component Analysis) for EEG signal analysis
4.Discussion and future work ICA (Independent Component Analysis) for EEG signal analysis Introduction Research Project Independent Component Analysis Frequency Domain EEG Data collection Methods and results Result Removing IC 2: Mavi Ruiz Blondet
Use the headset Emotiv Epoc to collect EEG signal data.
One of them is neutral signal (subject relaxed, without blinking), and the others are the mixed signal which regarded as the combination of neural signal and blinking signal. Each one has 678 sample points by time series.
The raw signal from the 14 channels acquired from the headset Emotiv Epoc:
We choose only the 4 channels affected by the artifacts. (the second picture on the right)
We use EEGLAB environment to do the ICA decomposition and component analysis. EEG signal How to use EEGLAB to do
the ICA decomposition and component analysis (ICA) is a statistical and computational technique for revealing hidden factors that underlie sets of random variables, measurements, or signals. Applications include:
1, Audio Processing(Cocktail party)
2, Biomedical Signals(EEG/MEG/MRI)
4, Imaging processing
5, Coding Principle 1: “Nonlinear decorrelation.
The components are uncorrelated, and the transformed components are uncorrelated.
Principle 2: “Maximum nongaussianity”.
Find the local maxima of nongaussianity of a linear combination y=Wx under the constraint that the variance of x is constant. After running the algorithm,we get 4 components:
(the third picture on the right) Here we can see the 4 components, 3 with the blinking and one with pure EEG data. Below we can see the properties of these components. The property which interests us the most is the frequency response, since these components will act as filters to separate the data. Removing IC2 and IC3: The final output, only eliminating IC2: Final output eliminating IC2 and IC3: Analysis Discussion Conclusion Amplitude: can range from 100µV to 10mV
Frequency: from 0.5Hz to 100Hz. Correlation Result shows that, in time domain, the Channels F7, F8, AF4 show higher correlation with neural data, if ICA being applied, along with higher information conservation (high correlation) of raw data.
In frequency domain AF3 and AF4 channel also shows relatively good result. Filters Pros and Cons of ICA ICA is not a tool designed for pulling small signals
out of (true) noise;
ICA assume the components are statistically linearly mixed, non-linear relationship being ignored in this model;
If the number of independent components is large, it will degrades the accuracy of ICA. Without knowing any relationship within the model, we still can separate independent source with high accuracy and reveal the hidden factors Magnitude and phase Fourier Transform For any signal And frequency domain was born Time domain is just x(t), can be plot as x(t)/t
Frequency domain is X(w), can be plot as amplitude and phase over the harmonic frequencies,(2 plots) FT: IFT: http://ay20-danield.blogspot.com/2012/10/fourier-transforms-lecture-response.html Another way of seeing the world: Discrete Fourier Transform IDFT: DFT: Artifacts EMG (clenching jaw)
Power supply of 60Hz
Potential related to cardial activity AF3 F7 AF4 F8 (The human eye can be considered a dipole that is positively charged in front and negatively behind.) Our gold standard The traditional approach to process signals and eliminate the noise is to filter the undesired frequencies.
It is very simple to implement a notch filter in Matlab. The filtered output will be compared to the ICA output to evaluate the results % Enter the frequency interval, in hertz, for filtering the data:
% Invoke an ideal notch filter:
idealfilter_countn = idealfilter(count1,interval,'notch') 1 1.15 4.25 Filter magnitude response Removing IC2, IC3 and IC1: Original
Components COCKTAIL PARTY PROBLEM Imagine you're at a cocktail party. For you it is no problem to follow the discussion of your neighbours, even if there are lots of other sound sources in the room: other discussions in English and in other languages, different kinds of music, etc.
It is not known exactly how humans are able to separate the different sound sources. Independent component analysis is able to do it, if there are at least as many microphones or 'ears' in the room as there are different simultaneous sound sources. In this demo, you can select which sounds are present in your cocktail party. ICA will separate them without knowing anything about the different sound sources or the positions of the microphones. Frequency Domain Time Domain Original data with blinks Original data with no blinks IC2 removed IC2 and IC3 removed Filtered data Hz There are many algorithms available to implement ICA, including fastICA
We can foreseen that it will become a standard method in the future.
ICA is a better choice than filter and there is less risk to lose information Correlation has been chosen as an analysis method, we use it on time domain as well as on frequency domain. We use Fourier Transform to achieve frequency domain analysis. Traditional filter technique has been used as parallel experiment to compare with the result from ICA method.