a design and implementation of an emotion reasoning chip ... · a design and implementation of an...
TRANSCRIPT
Abstract—The implementation of the emotion reasoning
systems to these days has been on software level; the software
evaluates emotional states based on the amount of changes in
biological signals. Therefore, the memory consumption and
computational overload would increase with the number of
parameters used, and the overall effect would be an increased
the response time needed to acquire the needed emotion
reasoning results. In this paper, we go over the design of a
One-Chip hardware reasoning block dedicated to emotion
based applications that are capable of processing multiple,
real-time information, and the design an emotion reasoning
system utilizing this chip. Our system provides an emotion
reasoning API for the interaction between sensing and
reasoning blocks so that the hardware emotion reasoning device
can be easily utilized by emotion based service applications. In
addition, it is possible to manufacture accessory devices for
smartphones based on downsized reasoning block hardware
with USB and Bluetooth interface for use with emotion based
mobile services.
Index Terms—Emotion reasoning, emotion inference.
I. INTRODUCTION
In order to acquire numerical values that can be used to
infer about individuals’ emotional states, sensors that
measure variables such as PPG (Photoplethysmogram),
GSR(Galvanic Skin Response), and SKT(Skin Temperature)
are used [1], [2]. However, individuals’ actual emotional
states are likely to differ even under identical conditions [3],
[4], and environmental factors such as temperature and
humidity affect individuals’ emotions differently. Therefore,
not only the biological data gathered from the mentioned
sensors but also additional factors that could influence
individuals’ emotions should be considered. Diverse types of
data along with the feedback from users’ unique emotional
states can help obtain a more truthful picture of human
emotional states.
An emotion reasoning system can be divided largely two
parts, a sensing block and a reasoning block. The sensing
block utilizes a few sensors to send biological signals
acquired from sensors for PPG, GSR, SKT, HR (heart rate),
etc. from the human body to the reasoning block. Depending
on whether the sensor output values have increased,
decreased, or remained the same, the reasoning block then
makes an evaluation of the current emotional state based on a
Manuscript received January 25, 2013; revised March 25, 2013. This
work was supported by the IT R&D program of MKE/KEIT.
[2009-S-014-01, On the development of Sensing based Emotive Service
Mobile Handheld Devices]
The authors are with the Wireless Convergence Platform Research Center,
Korea Electronics Technology Institute, Seoul, Korea (e-mail: busytom@
keti.re.kr, [email protected], [email protected], [email protected],
rule-base table. Hence, the number of entries in the rule-base
table in the reasoning block increases with the number of
parameters needed for making decision, and this degrades the
computational speed.
In this paper, we discuss the design and implementation for
an emotion reasoning system based on One-Chip integral
emotion reasoning hardware. By constructing the reasoning
block on a single hardware device, one can achieve lower
memory consumption and faster computational speed. Since
the one-chip emotion reasoning hardware can be refined for
use in smartphones and smartpads as an accessory device,
this design can form the basis for diversification of emotion
based contents.
II. EMOTION REASONING
In emotion reasoning systems, the sensing block collects
biological data and transmits to the reasoning block. In
software-based systems, the reasoning block receives the
data collected in the sensing block, processes it as necessary,
and defines the current emotional state as one of the
following: pleasant, unpleasant, aroused, relaxed, pleasantly
aroused, pleasantly-relaxed, unpleasantly aroused,
unpleasantly-relaxed, or neutral. Additionally, parameters
such as environmental temperature that could affect
emotional states can be used to further define emotional
Fig. 1. Emotion reasoning results.
Fig. 2. Basic rule-base table.
For emotion reasoning, a rule-based table [6], [7]
pertaining to sensed data value decrease or increase is used.
Say there are four parameters being used, namely PPG, GSR,
A Design and Implementation of an Emotion Reasoning
Chip Based Emotion Reasoning System
Yong-Seok Lim, Seung-Ok Lim, Young-Choong Park, Byoung-Ha Park, and Joongjin Kook
131DOI: 10.7763/LNSE.2013.V1.29
states [5]. The final resulting evaluation is shown in Fig. 1.
Lecture Notes on Software Engineering, Vol. 1, No. 2, May 2013
Fig. 3. Compound rule-base table.
sin _ 3sen g paramsThe Number of Entries Pos Time THI
(1)
When each field uses 1 byte of memory, one entry takes up
8 bytes and the size of the entire table is 11644bytes
(11.4KB). The price of including the rule-base table into the
emotion based application program is an increase in the size
of the program and a drop in the computational speed.
III. ONE-CHIP HARDWARE BASED EMOTION REASONING
SYSTEM
Fig. 4. Emotion reasoning system architecture.
The computational overload of an emotion reasoning
system is lower when the reasoning block, the essential
component of the system, is made into hardware instead; the
emotion-based application can use an API to send the
sensor-measured data to the reasoning block and then receive
the result from the reasoning block. Moreover, since the
Fig. 5. USB-type one-chip emotion reasoning device.
However, since Android smartphones/pads released in the
market so far do not officially support devices other than
popular USB-based products such as keyboards, mouses,
memory sticks, it is necessary to modify the kernel to run
hardware based emotion reasoning system. Galaxy S2 and S3,
the smartphones used in this paper, use a customized Linux
kernel source from Samsung Electronics. One can use the
One-Chip emotion reasoning device on these models by
modifying the kernel setting so that it provides USB function
Fig. 6. Kernel setting for USB-to-Serial communication.
IV. EMOTION REASONING DEVICE AND APIS
An emotion based application should have the capability to
collect and process data from an external sensing device. The
application should also periodically send a query containing
the processed data and additional information to the emotion
132
SKT, and HR. Each parameter can have three possible stored
values: increase (+1), decrease (-1), no change (0). Such
rule-base table can be laid out as Fig. 2. Depending on how
much information is being considered for emotion reasoning,
the number of entries in the table can increase dramatically as
shown in Fig. 3.
Newly introduced parameters in Fig. 3 are position, time,
and temperature. States of position is categorized as either
indoor or outdoor, and of time as AM, PM, and late night.
States of temperature is categorized into level 1, 2, and 3. The
number of entries in the rule-base table is calculated as
follows:
rule-base table is included in the reasoning block, the size of
the emotion-based application minimized as well. Our
One-Chip hardware based reasoning device has a microUSB
interface so that it can be easily utilized by smartphones or
smartpads. The overall structure is shown in Fig. 4. In Fig. 4,
emotion related applications use Bluetooth to receive data
from the sensing device and communicates with the
reasoning block using USB or Bluetooth.
An USB-type emotion reasoning device can be connected
to the smartphones/pads in the form of an USB accessory,
and a Bluetooth emotion reasoning device can be connected
to the smartphones/pads wirelessly. Fig. 5 shows the
USB-type emotion reasoning device.
needed for communication. Fig. 6 shows the result of
modifying the kernel setting for USB-to-Serial
communication.
Lecture Notes on Software Engineering, Vol. 1, No. 2, May 2013
reasoning engine. The application should receive the result
from the engine when the computation is complete. These
functions realized as an emotion API make it possible for
emotion based applications to utilize the reasoning engine.
TABLE I: EMOTION REASONING APIS
Function Description
BioDevConfig() Configure Bluetooth for use with sensing
device
BioDevConnect() Connect using Bluetooth with sensing
device
BioDevDisconnect() Disconnect Bluetooth connection to
sensing device
BioDevRecv() Receive biological signal
EREngInit() Initialize emotion reasoning engine
EREngConnect() Connect emotion reasoning engine
EREngDisconnect() Disconnect emotion reasoning engine
EREngSendQuery() Send query for emotion reasoning
EREngRecvResult() Receive emotion reasoning result
EREngSendERTable() Send rule-base table
EREngUpdateERTable() Update rule-base table
In order to test our emotion reasoning system’s operations,
one needs a sensing device to measure and transfer biological
signals to the application. The application needs to convert
the sensing data into the format required by the emotion
reasoning engine before sending it to the engine, and the
engine in return sends the emotion results to the application.
In order to process such procedures, we can use the emotion
API laid out in Table I. The general procedure for emotion
reasoning is shown in “Fig. 7”.
Fig. 7. Emotion reasoning procedure.
Fig. 8. Emotion reasoning system make up and Android applications.
“Fig. 8” and “Fig. 9” show the realized the emotion
reasoning system using the One-Chip emotion reasoning
hardware device and emotion API, and the demonstration of
emotion reasoning.
In case of an emotion reasoning system realized purely on
software level on Galaxy S2 and S3, the CPU occupancy
level is around 24.7%. One-Chip hardware reasoning block,
on the other hand, shows an occupancy level of around 5.3%.
Using software-based emotion reasoning applications on
Smartphones/pads that support multitasking not only causes a
significant burden on the system but also shortens battery life,
defeating the purpose of commercialization.
Fig. 9. Emotion reasoning results using one-chip emotion reasoning device
and emotion API.
V. CONCLUSION
Emotion reasoning systems can be divided into two parts, a
sensing block and a reasoning block, and the sensing block
collects and transfers data to the reasoning block. In this
scheme, the size of the rule-base table increases with the
number of parameters and this decreases the computational
speed. In this paper, we showed how the reasoning block can
be separated from the application by utilizing a One-Chip
hardware emotion reasoning device so that the memory
consumption and computation are kept low. The hardware
reasoning device has USB and Bluetooth interface,
facilitating its use as an accessory device in
smartphones/pads. This type of device can be widely applied
to mobile applications targeted at emotion related services.
ACKNOWLEDGMENT
This work was supported by the IT R&D program of
MKE/KEIT. [2009-S-014-01, On the development of
Sensing based Emotive Service Mobile Handheld Devices]
REFERENCES
[1] R. W. Picard, Affective Computing, Cambridge, MA: The MIT Press,
1997.
[2] N. Sebe, I. Cohen, T. Gevers, and T. S. Huang, “Multimodal
approaches for emotion recognition: a survey,” in Proc. the SPIE:
Internet Imaging, 2004, pp. 56-67.
[3] M. Whang, “The Emotional Computer Adaptive to Human Emotion,”
Probing Experience, Philips Research Book Series, vol. 8, pp. 209-219,
2008.
[4] M. Whang and J. Lim, “A Physiological Approach to Affective
Computing,” Affective Computing: Focus on Emotion Expression,
Synthesis and Recognition, In-Tech Education and Publishing, pp.
309-318, 2008.
[5] Y. Park, B. Park, K. Choi, K. Jung, and S. Kim, “A Development of
Human Emotion Interaction Platform in Smart Environments,” in Proc.
the 2nd International Conference onComputers, Networks,
Systems,and Industrial Applications, 2012, pp. 571-575.
[6] M. Black and Y. Yacoob, “Tracking and recognizing rigid and
non-rigid facial motions using local parametric models of image
motion,” in Proc. International Conference on Computer Vision, 1995,
pp. 374-381.
[7] Y. Yacoob and L. Davis, “Recognizing human facial expressions from
long image sequences using optical flow,” IEEE Transactions on
Pattern Analysis and Machine Intelligence, vol. 18, no. 6, pp. 636-642,
1996.
133
Lecture Notes on Software Engineering, Vol. 1, No. 2, May 2013
134
Yong-Seok Lim received his B.S. degree in Electronics
Engineering from Korea University, South Korea, in
2001 and his M.S. degree in Ultra Large Scale
Integration Digital System from Korea University in
2003. He is currently a senior researcher of Wireless
Convergence Platform Research Center of the Korea
Electronics Technology Institute. His areas of expertise
include wireless/network SOC system architecture and
wireless power transfer.
Young-Choong Park received his B.S., M.S., and
Ph.D. degrees in Computer Engineering from SeJong
University, South Korea, in 1999, 2001, and 2010. He is
currently a senior researcher of Realistic Media
Platform Research Center of the Korea Electronics
Technology Institute. His areas of expertise include
embedded software, wireless sensor network,
smart-learning, and mobile platforms.
Byoung-Ha Park received his B.S. and M.S. degrees
in Computer Engineering from SeJong University,
South Korea, in 1999, and 2001. He is currently a
senior researcher of Realistic Media Platform Research
Center of the Korea Electronics Technology Institute.
His areas of expertise include embedded software,
wireless sensor network, smart-learning, and mobile
platforms.
Joongjin Kook received his B.S. and M.S. degrees in
Computer Engineering from KwangWoon University,
South Korea, in 2005, and 2007, and his Ph.D. degree
in School of Computer from Soongsil University in
2012. He is currently a Post-Doc. of Realistic Media
Platform Research Center of the Korea Electronics
Technology Institute. His areas of expertise include
embedded software, wireless sensor network,
smart-learning, and mobile platforms.
Seung-Ok Lim respectively received B.S., M.S., and
Ph.D. degrees in Electrical Engineering from KonKuK
University, Seoul, Korea, in 1997, 1999, and 2005. He
is now a director at the Wireless Convergence Platform
Research Center in KETI (Korea Electronics
Technology Institute), Seoul, Korea. His research
interests include wireless communication systems for
harsh environments such as underground and
underwater environments, and wireless power transfer systems based on
magnetic coupling mechanism.
Lecture Notes on Software Engineering, Vol. 1, No. 2, May 2013