Biomedical Engineering Reference
In-Depth Information
×
10
4
EEG2 (2048x 1 real, Fs=250)
1
0.5
0
−
0.5
−
1
−
1.5
0
1
2
3
4
5
6
7
8
Time
Figure 3.10
EEG signal obtained from a rat model after brain injury.
10
4
Power spectra
×
16
spect1:Welch:Nfft=1024
spect1:Welch:Nfft=1024
14
12
10
8
6
4
2
0
0
20
40
60
80
100
120
Frequency
Figure 3.11
Comparison of the PSDs of EEG signals obtained from a rat model before (spect1) and
after brain injury (spect2).
states that arbitrary good time and frequency resolutions at the same location
cannot be achieved [47]:
1
4
ΔΔ
ft
≥
(3.24)
π
t
are the frequency and time resolutions, respectively. Nevertheless,
the time-frequency representation of the EEG signal will provide a better alternative
than having EEG information in either the time or frequency domain. Several tech-
niques have been proposed to solve this problem. We will describe the short-time
Fourier transform (STFT) and the wavelet transform (WT).
where
Δ
f
and
Δ
3.1.3.1 Short-Time Fourier Transform
The starting point with the STFT is to slice the EEG signal into short “stationary”
segments. This is usually performed by multiplying the EEG signal with a slid-
Search WWH ::
Custom Search