Professional Documents
Culture Documents
I. INTRODUCTION
Brain robot interaction (BRI) technology allows an
individual, such as a disable, to control the robotic devices via
brain signals [1][2][3]. Brain-controlled systems, based on
brain-computer interface (BCI), are classified into two
categories: the invasive technology implanting electrodes over
brain cortex through surgery to record signals of neurons, such
as Electrocorticogram (ECoG); the non-invasive technology
acquiring brain signals through electrodes placed on scalp,
Manuscript received July 24, 2015. The work is supported in part by The
National Natural Science Foundation of China (No. 61473207) and State Key
Laboratory of Robotics at Shenyang Institute of Automation (Grant No.
2014-Z03).
Danyang Zhi is with the School of Electrical Engineering and
Automation, Tianjin University, Tianjin 300072, China. (e-mail:
zhidanyang@tju.edu.cn).
Zhengping Wu is with the School of engineering education, Sanjiang
University, Nanjing 210000, China. (e-mail: wzp_959@163.com )
Wei Li is with the School of Electrical Engineering and Automation,
Tianjin University, Tianjin 300072, China. Shenyang Institute of Automation,
Chinese Academy of Sciences, Shenyang 110016, China. Department of
Computer & Electrical Engineering and Science, California State University,
Bakersfield, CA 93311, USA. Wei Li is the author to whom correspondence
should be addressed. (phone: +661-654-6747; fax: +661-654-6960; e-mail:
wli@csub.edu).
Jing Zhao is with the School of Electrical Engineering and Automation,
Tianjin
University,
Tianjin
300072,
China.
(e-mail:
zhaoj379779967@163.com).
Xiaoqiang Mao is with the School of Electrical Engineering and
Automation, Tianjin University, Tianjin 300072, China. (email:
maoxiaoqian@tju.edu.cn)
Mengfan Li is with the School of Electrical Engineering and Automation,
Tianjin University, Tianjin 300072, China. (e-mail: shelldream@ tju.edu.cn ).
Meifang Ma is with the Brain-Medical Ltd, Nantong 226300, China.
(e-mail: ediotmmf@gmail.com )
1018
(a)
(b)
Figure 2. (a) Diagram for acquisition module (b) Diagram for data
processing module
(a)
(a)
Figure 3.
(b)
(b)
1020
=1
,
sin(2 + , ) + ,
= 1,2, , ,
(1)
TABLE I.
subjects
Frequency-domain characteristics
systems
The
portable
system
Cerebot
S1
S2
S3
S4
Average
4.61Hz
31.2
19.2
17.9
19.1
21.9
20Hz
13.6
17.1
4.7
11.9
11.8
15Hz
7.2
17.1
5.4
6.9
9.2
4.61Hz
25.9
16.6
19.8
11.8
18.5
20Hz
17.2
8.7
6.7
16.9
12.4
15Hz
10.4
8.9
6.7
12.2
9.6
V. EXPERIMENTAL RESULT
A. Experimental Preparation
In order to evaluate the system performance, we off-line
analyze the brain signals and on-line control the Lego robot by
comparing the developed brain-controlled robot system with
Cerebot [13][14][15] because our previous studies via a variety
of brain models [15][26], including both the off-line analysis
and on-line control, demonstrated that Cerebot achieved high
accuracy and reliability of controlling a humanoid robot.
The Cerebot system is based on a Cerebus data acquisition
system [27]: a professional multichannel acquisition system
used for bio-potential signals recording, processing and
displaying. The system dimension is approximate 0.123 and
weight is about 7.4kg. The Cerebus system, consisting of the
signal amplifier, the amplifier power and the neural signal
processor, can record simultaneously brain signals from 128
channels at a sampling rate of 30 kHz and with 16-bits
resolution [15].
Four subjects with normal or corrected to normal vision,
aged among 23 to 27 years old, participated in the comparative
studies. The subjects sat in front of the monitor with a distance
away from about 60cm. For the two brain-controlled robot
systems, we used a 64-channels standard electrode cap at the
same sample frequency of 500Hz. The two systems adopted
the same off-line analysis and on-line control modules except
the data collecting and recording procedures. The experimental
process has been examined by Tianjin medical university
general hospital ethical review committee, and all subjects
have signed a written informed consent before the experiment.
B. Off-line Analysis
For the off-line analysis, the subjects were required to
conduct 90 trails to collect samples data using the two systems
1021
TABLE II.
CLASSIFICATION ACCURACY /%
subjects
systems
S1
S2
S3
S4
Average
The portable
system
Cerebot
96.7
95.6
92.2
91.1
93.9
Off-line
96.7
95.6
91.1
98.9
95.6
93.3
94.4
92.2
85.6
91.4
On-line
The portable
system
Cerebot
93.3
95.6
90.0
88.9
92.0
(a)
(b)
(c)
Figure 6.
(a) Live video feedback (b) Scene (c) The moving of Lego
1022
TABLE III.
RESULT OF TASK
Subjects
Experiment
Command Num.
First
Collision Num.
S1
S2
S3
S4
17
27
22
23
Command Num.
18
23
22
26
Collision Num.
Command Num.
18
23
28
19
Collision Num.
Second
Third
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
VII. CONCLUSION
[18]
[19]
[20]
[2]
[22]
[23]
[24]
[25]
REFERENCES
[1]
[21]
[26]
[27]
1023