Brain signal seminar

1. AN EEG-BASED BRAIN COMPUTER
INTERFACE FOR
EMOTION RECOGNITION AND ITS
APPLICATION IN
PATIENTS WITH DISORDER OF
CONSCIOUSNESS
HAIYUN HUANG, QIUYOU XIE, JIAHUI PAN, YANBIN HE, ZHENFU WEN,
RONGHAO YU AND YUANQING LI,
2019
PRESENTED BY
NAFIZ ISHTIAQUE AHMED
23-JANUARY-2020
UOU- UNIVERSITY OF ULSAN
Computational
Neural
Engineering Lab.

2. DISORDER OF CONSCIOUSNESS (DOC)
 Brain Injuries :
 Coma,
 Vegetative state (VS),
 Minimal conscious state
(MCS),
 Emergence from MCS (EMCS)

3. INTRODUCTION
 Emotion has strong connection with Consciousness.
 If DOC patients emotion is recognized then it can help to
assess their residual consciousness as well as their impaired
brain functions.
 As DOC patients suffers from severe motor impairments and
unable to provide proper emotion expressions; So far doctors
cannot detect their emotional states.

4. EMOTION RECOGNITION
Evoked Mechanisms Accuracy
Viewing Picture EEG 80.77%
listening to music EEG 82.29% ± 3.06%
Watching videos EEG, pupillary responses,
and gaze distances
76.4%
Watching videos EEG 79.28%
listening to music
(Real-time)
EEG 53.96%
 Emotion Recognition of Healthy people.

5. INTRODUCTION
 EEG is significant and widely used for identifying
human emotion state.
 This paper shows EEG Based BCI system to
recognize
the emotions of DOC patients at real-time.
 Emotion evoked by 2 class video clips
 Positive
 Negative

6. SUBJECTS
 Control group
(validate the BCI system)
 10 student (8-male, 2-female)
 Mean age 26
 Normal vision and hearing
 DOC Patient
(applied the BCI system)
 8 patient (6-male, 2-female)
 Mean age 35
 Stable condition with normal vision
and hearing
 No psychiatric medications (2-days)
 Clinical diagnosis
 2 patient-VS,
 5 patient-MSC,
 1 patient-EMCS

7. STIMULUS
 Initially, 140 Chinese movie clips (30s) that contained positive
or negative scenes was collected.
 10 volunteers evaluate their emotions with a level (i.e., not at
all, slightly, or extremely) and a keywords (i.e., positive or
negative) while watching the clips.
 Finally, 40 Chinese video clips(20 positive ,20 negative ) that all
volunteers scored as extremely positive or negative were
selected.
 Only 2 emotional state is chosen because complex and many
emotional states may increase the burden on the patients.

8. EXPERIMRNT
 SVM model (20 trials-train, 10-positive, 10-negative, 20
trails-test)

9. DATA ANALYSIS
 Baseline corrected by subtracting the mean value of the 1s
signal before the stimulus start.
 Notch filter was applied to remove the 50 Hz power-line noise.
 Tenth order minimum-phase FIR bandpass filter between 0.1 to
70 Hz.
 Online - Spectral power – STFT - a non-overlapped Hanning
window of 1 second- band power values are calculated by
averaging the power values in each frequency bands -
logarithmic scale – SVM model – Prediction.
 Offline - preprocessing, feature extraction -classification
procedures are the same as online method - 10 times 5-fold
cross-validation.

10. RESULTS (HEALTHY SUBJECT)

11. RESULTS (HEALTHY SUBJECT)

12. Topographical maps of the classification weight of each electrode :
average of the weights of all five subbands
1. The left frontal areas correlated to
positive emotion.
2. The right hemisphere mainly processed
negative emotion .
3. The reported frontal midline areas were
associated with the process of positive
emotion.

13. Topographies of different frequency bands
1. Depicts the average power changes for negative
and positive emotions in the five bands (delta,
theta, alpha, beta, and gamma).
2. In the delta band, the right anterior areas were
activated more for positive emotion than for
negative emotion.
3. In the theta band, the prefrontal regions and
occipital lobe show higher power during positive
emotional state than during negative emotional
state.
4. In the alpha band, the power decreased in the right
frontal areas during negative emotion, the power of
the frontal areas increased during positive emotion.
5. In the beta and gamma bands, the power in the
lateral temporal areas for positive emotion was
significantly higher than that for negative emotion.

14. Results for Patients with
DOC:

15. Results for Patients with
DOC:

16. CONCLUSION
1. An EEG-based BCI system to distinguish video-induced positive
and negative emotions.
2. Positive & Negative emotions were well evoked and recognized
by this BCI system.
3. This system provides an potential approach to detect the
emotions in patients with DOC.
4. The emotion BCI system may be a potential tool for evaluating
the consciousness levels of patients with DOC.