IRJET- Emotion Recognition using Facial Expression

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 04 | Apr 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1833
Emotion Recognition using Facial Expression
Aman Agrawal1, Anubhav Bhardwaj2, Prof. Sanjeev Kumar3
1,2,3Computer Science and Engineering, ABES Institute of Technology, Ghaziabad, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Facial Expressions are themostimmediatemode
of communication and interaction between individuals for
sharing info between themselves. Although Facial Emotion
Recognition can be conducted using multiple sensors but this
review focuses on studies that exclusively use facial images,
because visual expressions are one of the main information
channels in interpersonal communication. Automaticemotion
detection using facial expressions recognition is now a main
area of interest within various fields. These do not disclose the
feelings of anyone or don’t disclose someone’s mental views at
a particular period of time. This paper basically includes the
introduction of the human emotion recognition using the
facial expressions of a particular human being at a particular
instance of time.
Key Words: Human Emotions, Facial Expressions,Inception
v3 model, Tenserflow, CK+ dataset.
1. INTRODUCTION
Human Facial Expressions plays a vital roleinday-to-daylife
communication between different individuals. By
recognizing the facial expression of a particularindividual,it
becomes easier for determining the basic human emotions
like anger, fear, disgust, sadness, happiness and surprise.
These expressions can vary in every individual. Facial
expressions are produced by the movement of different
parts of our faces at different conditions. The preliminary
observation of the mentality is primarily based on the most
common forms of facial recognition to distinguish and
recognize face talks. In general,SVM,LBPandGaborareused
to distinguish and recognize transparentfacebasedonHaar,
Adaboost and Neural Networks. For example, Kobayashi et
al. has made the isolation and recognition of the first state-
based neural network. Caifeng Shan et al. uses a SVM
dictionary based on LBP's activities to get a visual
recognition.
In recent years, as a new recognition method that combines
artificial neural network and deep learning theory, the
convolutional neural network has made great strides in the
field of image classification. This method uses the local
reception field, the weight of the exchange and exchange
technology, and significantly reduces training parameters
compared to the neural network. It also has some degree of
translational invariance,rotationanddistortionoftheimage.
It has been widely used in speech recognition, facial
recognition. Tensorflow is the second artificial intelligence
research and development system developed by Google to
support CNN (3N), the neural network (RNN) and another
model of neural network. Depth This system is widely used
in Google products and services. It has been used in more
than 100 automated learning projects, with more than a
dozen fields of speech recognition, artificial vision and
robotics.
1.1. CATEGORIZING FACIAL EXPRESSIONS & IT’S
FEATURES:
Facial expression presents a key mechanism for describing
human emotion. From the beginning to the end of the day,
humanity changes many emotions, it may be due to your
mental or physical circumstances. Although human beings
are full of various emotions, modern psychology defines six
basic facial expressions: happiness, sadness, surprise, fear,
disgust and anger as universal emotions[2].Themovements
of facial muscles help identify human emotions. The basic
facial features are the eyebrows, the mouth,the noseandthe
eyes. We show that it is possible to differentiate faces with a
prototypical expression from the neutral expression.
Moreover, we can achieve this with data that has been
massively reduced in size: in the best case the original
images are reduced to just 5 components. We also
investigate the effect size on face images, a concept which
has not been reported previously on faces. Thisenablesusto
identify those areas of the face that are involved in the
production of a facial expression. Fear, disgust ,anger,
surprise, happiness and sadness all are the differenttypes of

facial emotions which a human being expresses through its
facial expressions whichcanbemadethroughthemovement
of faces .These movement of faces i.e. making facial
expressions brings different emotions at the face of the
human being .
2. Related work:
In 1971, Paul from a psychological point of view Six basic
emotions (happiness, sadness, anger, disgust, astonishment,
fear) through cultures. In 1978, Ekman et al. [22] A facial
action was developed Encoding system (FACS) to describe
facial expressions. Today,itisastatementoffacialexpression
Work is based on previous work, this document They have
also selected six basic emotions and neutral emotions facial
expression classification. Lu Guanming et al. [2] proposed a
convolution neural network for recognition of facial
expressions, and failure strategy and expansion strategy for
the data set adopted to solve the problem of insufficient
training data and the problem of overfitting. C. Shan et al.
applications LBP and SVM algorithm to classify facial
expressions with an accuracy of 95.10%. Andre Teixeira
Lopes and others [7] used a deep convolutional neural
network to classify facial expression, accuracy reached
97.81%. In traditional layout and learning in order to ensure
the accuracy and reliability ofthe classification trained,must
meet two basic assumptions: first the training samples and
test samples are taken independent distribution; the second
is that it is necessary training data. In many cases, however,
these two the conditions are hard to find. The most likely
scenario is that the training data is outdated. This usually
requires label a large number of training data to needs of our
training, but it is very expensive to label new data, which
requires a lot of manpower and material resources.
The purpose of the transfer of education is apply the
knowledge of an environment to a new environment.
Compared to the traditional learning machine, transfer
learning relax requirements of the two basicassumptions,it
cannot require a large amount of training data, or only a
small amount of data can be labeled. Transfer of learning
makes. It does not have to be like traditional machine
learning. The training samples and test samples should be
independently and identically distributed. In the same
compared to the traditional network with initialization, the
learning speed of the transfer learning is much faster.
Proposed System:
This experiment is based on the Inception V3 model[8]. An
example of platform tensorflow [1], hardware core, Intel i7
at 2.9 GHz, 8 GB DDR3 at 1600 MHz. CK + dataset used for
experimental expression data sets (Cohn-Kanade [15], CK +
dataset). We can choose the 1004 images of the facial image,
which contains 7 basic shapes, happiness (158), Tris (155)
out of (103), horror (146) surprise (161), fear (137) and
neutral ( 144).
re-process the image. First, the inception v3 model [8] is the
image in jpg or jpeg format for the formation and format of
the image of the dataset for the PNG format, so that the
image format is converted to PNG format in JPG format
converted. Secondly, since the image of the dataset is the
image of the facial expression taken by the digital camera
and some of the color images and some images in gray, the
image must be converted into the image in grayscale. In
order to eliminate the interference of the floor and hair and
to improve the accuracy of the classification of the images,
we must cut off the area of the face of the image and use the
clipping image for its formation, verificationandtesting.The
Inception v3 network model [8]isa verycomplexnetwork.If
you train the model directly and the data from the CK +
record [12] is relatively small, it takes a long time and the
training data is insufficient. Therefore, we use transfer-
learning technology to recycle the Inception v3 model. The
last layer of the Inception v3 model [8] is a softmax classifier
because the Image Net dataset contains 1000 classes, so the
classifier has 1000 root nodes in the original network. Here
we need to remove the last layer of the network, set the
number of output nodes to 7 (the number of facial
expressions), and then reuse the network model. The last
layer of the model is formed by the later propagation
algorithm, and the Cross-Entropy Cost function is used to

adjust the weight parameter by computing the error
between the output of the softmaxlayerandthevectorofthe
given example category day.
RESULT:
The data used in the experiment werecomposedoffiveadult
human volunteers (three men and two women) and were
used to construct the system, recording each person twice
while watching a 210-second video. The video contained six
different scenarios (relaxation, humor, sadness, fear and
discussion) and experienced the most feelings from
FeelTrace; we noticed that every volunteer had a different
reaction to every part of the video. The video was collected
from YouTube and also contained audio,whichhada greater
effect on the participants. The face video of each participant
was collected from the FPGA Xilinx Spartan-6 LX45 and a
camera sensor connected to it, and then an HDMI output
from the FPGA was later connected to the computer; the
Opencv Acquisition Toolbox that is used to receiveandstore
the real-time video capabilities of the FPGA board for use in
the machine. The videos were in RGB colors and for each
color component of a pixel in a frame. Each participant was
tied from the shoulder with his face to the camera while
sitting in the chair. It is made to be natural and real;
therefore, the videos were recorded in the office with a
natural background and in some cases some people can be
seen walking in the frame of the camera. The 10 recorded
videos contain no audio, because it contains no paper; this
also reduced the volume of data and thereforethedata could
be processed faster. It was interesting that the volunteers
had a different reaction to each part of the video . So in
general the data collection contained 63,000 labeled
samples.
CONCLUSION:
This article deals with programs to see face-to-face and
various research challenges. Basically, these programs
include face recognition, resource extraction, and
classification. Differentapproachescanbeusedtodetermine
the level of better recognition. Highly respected standards
are effective. These methods provide a practical solution to
the visual impairment problem and can work well in a
restricted area. Emotional sensitivity to the state of thestate
is a complex problem and creates difficulties due to the
physical and psychological characteristics associated with
individual characteristics. Therefore, the research in this
field will continue to study for many years, as it is important
to solve many problems to create a user interface and better
recognition of the complex emotional conditions required.
REFERENCES
[1] Martín Abadi, Ashish Agarwal, et al: TensorFlow: Large-
Scale Machine Learning on Heterogeneous Distributed
Systems. CoRR abs/1603.04467 (2016)
[2] G. Donato, M.S. Bartlett, J.C. Hager, P. Ekman, T.J.
Sejnowski, “Classifying Facial Actions”, IEEE Trans.
Pattern Analysis and Machine Intelligence, Vol. 21, No.
10, pp. 974-989, 1999
[3] L.Torres, J. Reutter, and L. Lorente, “The importance of
the color information in face recognition,” in
Proceedings IEEE International Conference on Image
Processing, vol. 3, pp.627–631, 1999.
[4] Andre Teixeira Lopes, Edilson de Aguiar, Thiago
Oliveira-Santos: A Facial ExpressionRecognitionSystem
UsingConvolutional Networks.SIBGRAPI2015:273-280
[5] E. Friesen and P. Ekman: “Facial action coding system: a
technique for the measurement of facial movement,”
Palo Alto, 1978.
[6] Ioan Buciu, Constantine Kotropoulos, Ioannis Pitas: ICA
and Gabor representation for facial expression
recognition. ICIP (2) 2003: 855-858
[7] Caifeng Shan, Tommaso Gritti: Learning Discriminative
LBP-Histogram Bins for Facial Expression Recognition.
BMVC 2008: 1-10
[8] Martín Abadi, Ashish Agarwal, et al: TensorFlow: Large-
Scale Machine Learning on Heterogeneous Distributed
Systems. CoRR abs/1603.04467 (2016)
[9] Sung-Oh Lee, Yong-Guk Kim, Gwi-Tae Park: Facial
Expression Recognition Based upon Gabor-Wavelets
Based Enhanced Fisher Model. ISCIS 2003: 490-496
[10] C. Shan, S. Gong, and P. W. McOwan: “Facial expression
recognition based on local binary patterns: A
comprehensive study,” Image and Vision

IRJET- Emotion Recognition using Facial Expression

More Related Content

What's hot (20)

Similar to IRJET- Emotion Recognition using Facial Expression (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Emotion Recognition using Facial Expression