Please enable JavaScript.
Coggle requires JavaScript to display documents.
Signal Analysis Methods - Coggle Diagram
Signal Analysis Methods
Coherence
Limitations
requires stationary data (not relevant to WTC)
does not consider phase relation
phase-amplitude relations are unclear
Wavelet Transform Coherence
Procedure
calculating the cross-correlation as a function
of frequency and time
computes frequency content with
wavelets
as basic functions
data should be cleaned of
noise
artefacts
Benefits
⬆️ spectral and temporal resolution
better suited for non-stationary signals such as those arising from physiological processes
providing power of specific frequencies at each time point
Characteristics
reveals frequency information through signal duration with
decreasing time resolution with lower frequencies
IMPORTANT for physiological data since frequency content varies over time (non-stationary)
sensitive to
signal power
capable of capturing out-of-phase relationships between brain activity
only provides
linear relationship
between 1 location and same location on other participant
symmetric output
Calculation
:one: Wavelets used to calculate Power Spectral Density (PSD) via Continuous Wavelet Transform (CWT)
:two: Compute Cross Spectral Density (CSD) by calculating cross-correlation of PSDs
:three: Compute Magnitude Squared Coherence (MSC) from CSD and PSD (analogous to WTC)
phase consistency at different frequencies
represented on frequency or
time-frequency
domain
Phase Synchrony
Phase Locking Value (PLV)
works well with EEG
high temporal resolution of EEG allows the phase dynamics to be analysed better
more datapoints are acquired
from the recording location which can be used to determine the phase.
Assumptions
instantaneous phase values are obtainable (through wavelet or Hilbert transform)
Limitations
Only able to display linear symmetric relationship between inputs
only suited to bivariate data
measure IBC through identifying existence of phase relations between input data
Circular Correlation Coefficient
based on circular covariance of differences between observed and mean phases
Assumptions
data is circular
circular distribution is uniform
Studies
Mortality Threat
Procedure
manipulation of mortality threat
dyadic model (43 dyads) tasked with completing competitive & cooperative button-pressing
EEG usage
Two 64-channel EEG systems used
placed following international 10-10 system
Sampling rate of 250Hz with impedance below 50kohms
Eye movements recorded using Electro-Oculogram channels
Preprocessing
use of MNE python software suite
Band-pass filtering (1-45 Hz)
Re-referencing to average of left and right mastoids
Independent component analysis to remove eye-movement related artifacts
Processing
CircStat toolbox in MATLAB
used to measure IBS
Causality / Information flow
Granger causality
Premises
cause precedes effect
knowledge of cause improves prediction of effect
ratio of variance of residuals of two autoregressive models, where one is the past of another signal
Features
Conditional Form
multivariate
tolerant of cases where signals x and y are dependent on third signal z
Standard Form
bivariate
asymmetric
able to represent linear relationship between input data
past of signal x predicts signal y better than past of Y
Limitations
questionable suitability for fMRI and fNIRS
Hemodynamic Response Function (HRF)
physiological response pattern that describes changes in blood flow, oxygenation, brain metabolic activity in response to neural activity
HRF varies across locations
has been disproven for fMRI
delay between neuronal activity and HRF response
GC timelag may exclude neural function or include perceived neural function and contaminants
Partial Domain Coherence