Lecture 4 - CNS Flashcards
(26 cards)
induced power
from non phase-locked signals
needs time frequency analysis
can be isolated by subtracting the ERP of each trial
change in amplitude by stimulus
evoked power
from phase-locked signals
easily analysed by ERPs
can either be onset of new oscillation
or phase-reset of current oscillation
by stimulus
power stays the same
plotting oscillations
following a circle
periodicity
two full cycles in a circle
zero to positive and negative to zero
phase
from start to end of wave
butterfly plots
for non-aligned / non phase-locked signals
power
= amplitude of a frequency band squared
always positive
the envelope, independent of time
connects peaks of signal
ERPs origin questions
can be the result of either single signal or oscillation
you cannot know
oscillation examples in nature
light sound water electromagnetism whales' tails
relationship between frequency and power in nature
1 / f
negatively correlated
when frequency is high, power is low and vice versa
frequency scaling
making frequencies more equal
as low frequencies tend to have more power
relative, based on change over time
Fourier’s theorem
every wave can be decomposed into sine waves
FFT
Fast Fourier Transformation
from time to frequency domain
amplitudes are respresented in histograms
x = frequency, y = power
iFFT
taking out frequencies in frequency domain, thereby changing time domain
to filter out noise
you only know based on theory which frequencies you can take out and which you can’t
non-stationary signals
oscillations that change over time
convolving
= mulitplying
Morlet wavelet
stationary oscillation that has been convolved with a Gaussian
looks like a wave within a normal distribution
fewer cycles
more temporal resolution
more cycles
better frequency resolution
Heisenberg’s uncertainty principle in cycles
if you want better temporal resolution (fewer cycles), your frequency resolution goes down
and vice versa
dot product
time series are represented as vectors of numbers
when they have the same length, they can be multiplied
product = dot product
dot product meaning
the further from zero, the more similar
if zero, the cancelled each other out
is relative to range of amplitudes
dot product in the signal
appears as its own wave
connecting all the positive peaks
complex wavelet
three-dimensional wave looks like a spiral has a real and an imaginary part convolved with a sine wave characterises both power as well as wave
