Instrumentation & Measurement Magazine 24-4 - 84

for performing MF assessment on single subject and single
task cases [13], but they start to struggle in multi-subject and
multi-task analysis. In order to overcome this issue, we propose
a model based on the available knowledge about EEG
frequency bands' behavior to track the development of MF in
maritime operators (Fig. 2).
The most common way to analyze different EEG frequency
bands is by means of power spectral density (PSD).
Since PSD is a frequency domain analysis, the EEG data needs
to be transposed to frequency domain. The most commonly
used algorithm for this domain conversion is the Fast Fourier
Transform (FFT). When calculating the FFT, a moving window
function is used to help fulfilling the periodicity requirement
of FFT and to have a time-frequency domain analysis by repeating
the frequency analysis every t seconds. It is a common
practice to have the window length big enough to accommodate
at least two cycles of the slowest frequency of interest.
For EEG data, the slowest frequency of interest is usually the
inferior limit of δ band, which leads to a window size of four
seconds.
Fig. 2. Flow chart for mental fatigue assessment.
selected to ensure that activity in relevant brain regions could
be investigated.
The data preprocessing approach is depicted in Fig. 2. The
EEG data was sampled at 128 Hz. Digital notch filters were
applied at 50Hz and 60Hz. Since EEG data is sensitive to physiological
and external artifacts, we decided to apply an artifact
reduction approach. We did not want to include extra sensors
such as EOG and EMG, which are used to remove, for example,
blinking and chewing artifacts, since we intended to have the
simplest sensor setup possible. We therefore decided to apply
a detrend technique that could help mitigate the effect of different
kinds of artifacts without additional sensors.
The detrend approach relies on approximating the signal
to a low-polynomial function and subtracting it from the original
data. In order to mitigate problems that this traditional
detrend approach can add to the resulting signal, we implemented
a robust detrend algorithm. This approach makes use
of a masking function to guide the application of the polynomial
approximation, avoiding the inclusion of big punctual
outliers in the detrend process. For additional details on the robust
detrend algorithm implementation, see [12]. In order to
help mitigate the remaining signal spikes caused by external
sources of artifacts that cannot be reduced by the robust detrending
approach, we applied a threshold function of 200 μV.
Finally, each data channel was then mean centered to remove
the sensor reading offset.
MF Assessment Approach
One of the main disadvantages of using data-driven approaches
such as neural networks is that they usually rely on
big amounts of data for training the classification algorithm.
In general, these models can only be as good as the data (and
time) used to train them. With limited amounts of EEG data, a
neural network algorithm such as a CNN can be trained well
84
For each four seconds of EEG data, we calculated the correspondent
PSD spectrum to obtain a temporal variation
of the energy spectrum. In order to calculate the PSD spectrum
from the frequency spectrum, we applied the multitaper
periodogram, which has been successfully used to replace
the traditional periodogram in EEG analysis. The multitaper
periodogram offers a more robust estimation of the PSD
by averaging the result of several statistically independent
periodogram obtained by filtering the original signal using orthogonal
bandpass filters. The total energy in each frequency
band can be obtained by integrating the PSD spectrum.
With the time-varying PSD in each frequency band we can
calculate an energy ratio to be used during the MF assessment
process. Based on the previous frequency bands discussion
and previous results from the MF assessment literature [14],
we opted to use the following PSD ratio:
PSD t 
ratio
()
PSD t PSD t
()


()
PSD t
()
(1)
To make the assessment system as simple and noninvasive
as possible, we applied only one of the frontal electrodes in the
analysis, AF3. Since the oscillatory behavior typical from EEG
signals can lead to a PSD ratio with excessive variance, we applied
a local average function of 16 time steps to smooth out
the response.
The physiological differences among people and specific
requirements, both physical and mental, of different tasks,
make it very unlikely that a unique MF scale can unequivocally
detect MF levels in any scenario. In order to overcome
this issue, we decided to evaluate how the MF progresses
during a task. The progression speed can be used as a better
indication of the increasing risk associated with an individual
performing a specific task. In order to be able to compare the
MF progression in the cross-subject and cross-task analysis, we
applied a normalization to the averaged PSD ratio. The normalizing
factor is based on the global average of the PSD ratio:
IEEE Instrumentation & Measurement Magazine
June 2021

Instrumentation & Measurement Magazine 24-4

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-4

No label
Instrumentation & Measurement Magazine 24-4 - No label
Instrumentation & Measurement Magazine 24-4 - Cover2
Instrumentation & Measurement Magazine 24-4 - 1
Instrumentation & Measurement Magazine 24-4 - 2
Instrumentation & Measurement Magazine 24-4 - 3
Instrumentation & Measurement Magazine 24-4 - 4
Instrumentation & Measurement Magazine 24-4 - 5
Instrumentation & Measurement Magazine 24-4 - 6
Instrumentation & Measurement Magazine 24-4 - 7
Instrumentation & Measurement Magazine 24-4 - 8
Instrumentation & Measurement Magazine 24-4 - 9
Instrumentation & Measurement Magazine 24-4 - 10
Instrumentation & Measurement Magazine 24-4 - 11
Instrumentation & Measurement Magazine 24-4 - 12
Instrumentation & Measurement Magazine 24-4 - 13
Instrumentation & Measurement Magazine 24-4 - 14
Instrumentation & Measurement Magazine 24-4 - 15
Instrumentation & Measurement Magazine 24-4 - 16
Instrumentation & Measurement Magazine 24-4 - 17
Instrumentation & Measurement Magazine 24-4 - 18
Instrumentation & Measurement Magazine 24-4 - 19
Instrumentation & Measurement Magazine 24-4 - 20
Instrumentation & Measurement Magazine 24-4 - 21
Instrumentation & Measurement Magazine 24-4 - 22
Instrumentation & Measurement Magazine 24-4 - 23
Instrumentation & Measurement Magazine 24-4 - 24
Instrumentation & Measurement Magazine 24-4 - 25
Instrumentation & Measurement Magazine 24-4 - 26
Instrumentation & Measurement Magazine 24-4 - 27
Instrumentation & Measurement Magazine 24-4 - 28
Instrumentation & Measurement Magazine 24-4 - 29
Instrumentation & Measurement Magazine 24-4 - 30
Instrumentation & Measurement Magazine 24-4 - 31
Instrumentation & Measurement Magazine 24-4 - 32
Instrumentation & Measurement Magazine 24-4 - 33
Instrumentation & Measurement Magazine 24-4 - 34
Instrumentation & Measurement Magazine 24-4 - 35
Instrumentation & Measurement Magazine 24-4 - 36
Instrumentation & Measurement Magazine 24-4 - 37
Instrumentation & Measurement Magazine 24-4 - 38
Instrumentation & Measurement Magazine 24-4 - 39
Instrumentation & Measurement Magazine 24-4 - 40
Instrumentation & Measurement Magazine 24-4 - 41
Instrumentation & Measurement Magazine 24-4 - 42
Instrumentation & Measurement Magazine 24-4 - 43
Instrumentation & Measurement Magazine 24-4 - 44
Instrumentation & Measurement Magazine 24-4 - 45
Instrumentation & Measurement Magazine 24-4 - 46
Instrumentation & Measurement Magazine 24-4 - 47
Instrumentation & Measurement Magazine 24-4 - 48
Instrumentation & Measurement Magazine 24-4 - 49
Instrumentation & Measurement Magazine 24-4 - 50
Instrumentation & Measurement Magazine 24-4 - 51
Instrumentation & Measurement Magazine 24-4 - 52
Instrumentation & Measurement Magazine 24-4 - 53
Instrumentation & Measurement Magazine 24-4 - 54
Instrumentation & Measurement Magazine 24-4 - 55
Instrumentation & Measurement Magazine 24-4 - 56
Instrumentation & Measurement Magazine 24-4 - 57
Instrumentation & Measurement Magazine 24-4 - 58
Instrumentation & Measurement Magazine 24-4 - 59
Instrumentation & Measurement Magazine 24-4 - 60
Instrumentation & Measurement Magazine 24-4 - 61
Instrumentation & Measurement Magazine 24-4 - 62
Instrumentation & Measurement Magazine 24-4 - 63
Instrumentation & Measurement Magazine 24-4 - 64
Instrumentation & Measurement Magazine 24-4 - 65
Instrumentation & Measurement Magazine 24-4 - 66
Instrumentation & Measurement Magazine 24-4 - 67
Instrumentation & Measurement Magazine 24-4 - 68
Instrumentation & Measurement Magazine 24-4 - 69
Instrumentation & Measurement Magazine 24-4 - 70
Instrumentation & Measurement Magazine 24-4 - 71
Instrumentation & Measurement Magazine 24-4 - 72
Instrumentation & Measurement Magazine 24-4 - 73
Instrumentation & Measurement Magazine 24-4 - 74
Instrumentation & Measurement Magazine 24-4 - 75
Instrumentation & Measurement Magazine 24-4 - 76
Instrumentation & Measurement Magazine 24-4 - 77
Instrumentation & Measurement Magazine 24-4 - 78
Instrumentation & Measurement Magazine 24-4 - 79
Instrumentation & Measurement Magazine 24-4 - 80
Instrumentation & Measurement Magazine 24-4 - 81
Instrumentation & Measurement Magazine 24-4 - 82
Instrumentation & Measurement Magazine 24-4 - 83
Instrumentation & Measurement Magazine 24-4 - 84
Instrumentation & Measurement Magazine 24-4 - 85
Instrumentation & Measurement Magazine 24-4 - 86
Instrumentation & Measurement Magazine 24-4 - 87
Instrumentation & Measurement Magazine 24-4 - 88
Instrumentation & Measurement Magazine 24-4 - 89
Instrumentation & Measurement Magazine 24-4 - 90
Instrumentation & Measurement Magazine 24-4 - 91
Instrumentation & Measurement Magazine 24-4 - 92
Instrumentation & Measurement Magazine 24-4 - 93
Instrumentation & Measurement Magazine 24-4 - 94
Instrumentation & Measurement Magazine 24-4 - 95
Instrumentation & Measurement Magazine 24-4 - 96
Instrumentation & Measurement Magazine 24-4 - 97
Instrumentation & Measurement Magazine 24-4 - 98
Instrumentation & Measurement Magazine 24-4 - 99
Instrumentation & Measurement Magazine 24-4 - 100
Instrumentation & Measurement Magazine 24-4 - 101
Instrumentation & Measurement Magazine 24-4 - 102
Instrumentation & Measurement Magazine 24-4 - 103
Instrumentation & Measurement Magazine 24-4 - 104
Instrumentation & Measurement Magazine 24-4 - 105
Instrumentation & Measurement Magazine 24-4 - 106
Instrumentation & Measurement Magazine 24-4 - 107
Instrumentation & Measurement Magazine 24-4 - 108
Instrumentation & Measurement Magazine 24-4 - 109
Instrumentation & Measurement Magazine 24-4 - 110
Instrumentation & Measurement Magazine 24-4 - 111
Instrumentation & Measurement Magazine 24-4 - 112
Instrumentation & Measurement Magazine 24-4 - 113
Instrumentation & Measurement Magazine 24-4 - 114
Instrumentation & Measurement Magazine 24-4 - 115
Instrumentation & Measurement Magazine 24-4 - 116
Instrumentation & Measurement Magazine 24-4 - 117
Instrumentation & Measurement Magazine 24-4 - 118
Instrumentation & Measurement Magazine 24-4 - 119
Instrumentation & Measurement Magazine 24-4 - 120
Instrumentation & Measurement Magazine 24-4 - Cover3
Instrumentation & Measurement Magazine 24-4 - Cover4
https://www.nxtbook.com/allen/iamm/24-8
https://www.nxtbook.com/allen/iamm/24-6
https://www.nxtbook.com/allen/iamm/24-5
https://www.nxtbook.com/allen/iamm/24-4
https://www.nxtbook.com/allen/iamm/24-3
https://www.nxtbook.com/allen/iamm/24-2
https://www.nxtbook.com/allen/iamm/24-1
https://www.nxtbook.com/allen/iamm/23-9
https://www.nxtbook.com/allen/iamm/23-8
https://www.nxtbook.com/allen/iamm/23-6
https://www.nxtbook.com/allen/iamm/23-5
https://www.nxtbook.com/allen/iamm/23-2
https://www.nxtbook.com/allen/iamm/23-3
https://www.nxtbook.com/allen/iamm/23-4
https://www.nxtbookmedia.com