Detecting Facial Expression in Images

Journal for Research | Volume 02 | Issue 02 | April 2016
ISSN: 2395-7549
Detecting Facial Expression in Images

Ghazaala Yasmin
M.Tech. Student
Department of Computer Science & Engineering
University of Calcutta
Prof. Samir K. Bandyopadhyay

Professor
Department of Computer Science & Engineering
University of Calcutta
Abstract
Now days, the task of face recognition is widely used application of image analysis as well as pattern recognition. In biometric
area of the research, automatically face & face expression recognition attracts researchers interest. For classifying facial
expressions into different categories, it is necessary to extract important facial features which contribute in identifying proper and
particular expressions. Recognition and classification of human facial expression by computer is an important issue to develop
automatic facial expression recognition system in vision community. In this paper the facial expression recognition system is
proposed.
Keywords: Facial feature detection, Template matching and Face position detection
_______________________________________________________________________________________________________
I.
INTRODUCTION
Facial expression recognition [1] [2] is another fruitfulness of computer vision research. Computer vision is the way to
electronically represent the human vision with the help of some data analysis techniques. In a human computer interaction (HCI)
system, the communication between human and computer can take place through different aspects (like verbal, non-verbal) of a
human. Here we are considering only the non-verbal aspects of human being like, facial expression, body movement etc. In this
paper we are mainly concern about the facial expression recognition process, which needs a face image on which we should
apply our facial expression recognition algorithm. Now once we have the face image data, we need to apply some processing
techniques with the help of pattern recognition, artificial intelligence, mathematics, computer science, electronics or any kind of
scientific concept. Hence there are huge numbers of applications in computer vision research, but we will discuss only face
recognition and facial expression.
There are many applications, where facial expression detection process plays an important role. Researches in the field of
social psychology show that facial expression are more natural in nature than the speakers spoken words and truly reflects the
emotion of a person. According to statistical reports verbal part of a message contributes only for 7 percent to the effect of the
message as a whole. The vocal part contributes for 38 percent, while facial expression of the speaker contributes for 55 percent
to the effect of the spoken message. The facial expression recognition system was introduced in 1978 by Suwa et. al [4]. The
main issue of building a facial expression recognition system is face detection [3] and alignment, image normalization, feature
extraction, and classification.
The analysis of the human face via image (and video) is one of the most interesting and focusing research topics in the last
years for the image community. From the analysis (sensing, computing, and perception) of face images, much information can be
extracted, such as the sex/gender, age, facial expression, emotion/temper, mentality/mental processes and behaviour/psychology,
and the health of the person captured. According to this information, many practical tasks can be performed and completed; these
include not only person identification or verification (face recognition), but also the estimation and/or determination of person's
profession, hobby, name (recovered from memory), etc.
Research on face image analysis has been carried out and is being conducted around various application topics, such as (in
alphabetical order) age estimation, biometrics, biomedical instrumentations, emotion assessment, face recognition, facial
expression classification, gender determination, human-computer/human-machine interaction, human behaviour and emotion
study, industrial automation, military service, psychosis judgment, security checking systems, social signal processing,
surveillance systems, sport training, tele-medicine service, etc.
Therefore facial expressions are the most important information for emotions perception in face to face communication. This
paper explains about an approach to the problem of facial feature extraction from a non-l frontal posed image For face portion
segmentation basic image processing operation like morphological dilation, erosion, reconstruction techniques with disk
structuring element are used. Six permanent Facial features like eyebrows(left and right), eye (left and right) , mouth and nose
are extracted using facial geometry, edge projection analysis and distance measure and feature vector is formed considering
height and width of left eye, height and width of left eyebrow, height and width of right eye, height and width of right eyebrow,
height and width of nose and height and width of mouth along with distance between left eye and eyebrow, distance between
right eye and eyebrow and distance between nose and mouth.
Human face detection has drawn considerable attention in the past decades as it is one of the fundamental problems in
computer vision. Given a single image, the ideal face detection should identify and locate all faces regardless of its threedimensional position, orientation, and lighting conditions. The existing face detection techniques can be classified into four
All rights reserved by www.journalforresearch.org
30

(J4R/ Volume 02 / Issue 02 / 006)
categories, namely, knowledge-based methods, feature invariant approaches, template matching methods, appearance based
methods.
Human face detection and segmentation is an active research area until recently. This field of research plays an important role
in many applications such as face identification system, face tracking, video surveillance and security control system, and human
computer interface.
Those applications often require segmented human face which is ready to be processed. There are many factors that influence
the success of human face detection and segmentation. Those factors include complex colour background, condition of
illumination, change of position and expression, rotation of head, and distance between camera and subject.
Face detection is a sub branch of object detection. The human face is a dynamic object and has a high degree of variability in
its appearance, which makes face detection a difficult problem in computer vision.
Images containing faces are essential to intelligent vision-based human computer interaction, and research efforts in face
processing include face recognition, face tracking, pose estimation, and expression recognition. However, many reported
methods assume that the faces in an image or an image sequence have been identified and localized. To build fully automated
systems that analyse the information contained in face images, robust and efficient face detection algorithms are required.
Given a single image, the goal of face detection is to identify all image regions which contain a face regardless of its threedimensional position, orientation, and lighting conditions. Such a problem is challenging because faces are non- rigid and have a
high degree of variability in size, shape, colour, and texture. Numerous techniques have been developed to detect faces in a
single image.
Face detection and localization is the task of checking whether the given input image contains any human face, and if so,
returning the location of the human face in the image. The wide variety of applications and the difficulty of face detection have
made it an interesting problem for the researchers in recent years.
Face detection is difficult mainly due to a large component of non-rigidity and textural differences among faces. The great
challenge for the face detection problem is the large number of factors that govern the problem space. The long list of these
factors include the pose, orientation, facial expressions, facial sizes found in the image, luminance conditions, occlusion,
structural components, gender, ethnicity of the subject, the scene and complexity of images background.
The scene in which the face is placed ranges from a simple uniform background to highly complex backgrounds. In the latter
case it is obviously more difficult to detect a face. Faces appear totally different under different lighting conditions. Not only do
different persons have different sized faces, faces closer to the camera appear larger than faces that are far away from the camera
Basic emotions are emotions that have been scientifically proven to have a certain facial expression associated with it. Different
emotional stages are indicated as follows.
31

(J4R/ Volume 02 / Issue 02 / 006)
Facial Feature Detection system is fast becoming a familiar feature in apps and on websites on different purposes. Human
face identification and detection is often the first step in applications such as video surveillance, human computer interface, and
face recognition and image database management [1]. Furthermore facial feature characteristics are very much effective in both
biometric identification which automatically identifies a person from a digital image or a video image. Facial expressions are
used not only to express our emotions, but also to provide important communicative cues during social interaction, such as our
level of interest [2]. Facial features, eye feature has more application domain. It is reported that facial expressions have
considerable effects on listener, near about 55 % effect of the spoken words depend on eye movements and facial expressions of
the speaker.
Facial expressions play a major role in Face Recognition Systems and image processing techniques of Human Machine
Interface. There are several techniques for facial features selection like Principal Component Analysis, Distance calculation
among face components, Template Matching. This algorithm describes a simple template matching based facial feature
selection technique and detects facial expressions based on distances between facial featuresusing a set of image databases. The
process for facial expression recognition system involves three stages: Pre Processing, Facial Feature Extraction and
classification facial expressions.
II. REVIEW WORKS
Human face detection has drawn considerable attention in the past decades as it is one of the fundamental problems in computer
vision. Given a single image, the ideal face detection should identify and locate all faces regardless of its three-dimensional
position, orientation, and lighting conditions. The existing face detection techniques can be classified into four categories,
namely, knowledge-based methods, feature invariant approaches, template matching methods, appearance based methods.
The use of colour information has been introduced to the face-locating problem in recent years. Most publications [1-5] have
shown that colour is a powerful descriptor that has practical use in the extraction of the face detection.
Modelling skin colour requires choosing an appropriate colour space and identifying a cluster associated with skin colour in
this space. YIQ colour space is used in commercial colour television broadcasting. YCbCr space is a hardware orientated model
and is used in most video standards. So an effective use of the chrominance information for modelling human skin colour can be
achieved in these colour space. Second, this format is typically used in video coding, and therefore the use of the same format for
segmentation will avoid the extra computation required in conversation. Many research studies [6- 8] assume that chrominance
components of skin-tone colour are independent of the luminance component. In fact, the skin-tone colour is non-linearly
dependent on luminance. Researcher found that although skin colours of different people appear to vary over a wide range, they
differ less in chrominance than brightness, specially the skin colours from a compact area in the YCbCr plane [9-10].
Human-like robots and machines that are expected to enjoy truly intelligent and transparent communications with human can
be created using automatic facial expression recognition with a set of specific desired accuracy and performance requirements.
Expression recognition involves a variety of subjects such as perceptual recognition, machine learning, affective computing etc.
One case study uses skin colour range of human face to localize face area. After face detection, various facial features are
identified by calculating the ratio of width of multiple regions in human face. Finally the test image is partitioned into a set of
32

(J4R/ Volume 02 / Issue 02 / 006)
sub-images and each of these sub-images is matched against a set of sub-pattern training set. Partitioning is done using AwSpPCA algorithm. Given as input any emotion of face, this pattern training set will classify the particular emotion [1].
Face component extraction by dividing the face region into eye pair and mouth region and measurement of Euclidean distance
among various facial features is also adopted by a case study. Similar study is done by Neha Gupta to detect emotions. This
research includes four steps: pre-processing, edge detection, feature extraction and distance measurement among the features to
classify different emotions. This type of approach is classified as Geometric Approach [3].
Another research includes Face detection method using segmentation technique. First, the face area of the test image is
detected using skin colour detection. RGB colour space is transformed into yCbCr colour space in the image and then skin
blocks quantization is done to detect skin colour blocks. As next step, a face cropping algorithm is used to localize the face
region. Then, different facial features are extracted using segmentation of each component region (eyes, nose, mouth). Finally,
vertical & angular distances between various facial features are measured and based on this any unique facial expression is
identified. This approach can be used in any biometric recognition system [2].
A template matching based facial feature detection technique is used in a different case study [4,7-8]. Different methods of
face detection and their comparative study are done in another review work. Face detection methods are divided into two primary
techniques: Feature based & View based methods [5, 9].
Gabor filters are used to extract facial features in another study. This approach is called Appearance based approach. This
classification based facial expression recognition method uses a bank of multilayer perceptron neural networks. Feature size
reduction is done by Principal Component Analysis (PCA) [6, 10]. Thus existing works primarily focused in detecting facial
features and they are served as input to emotion recognition algorithm. In this study, a template based feature detection technique
is used for facial feature selection and then distance between eye and mouth regions is measured.
III. FACIAL EXPRESSION RECOGNITION SYSTEM AND RESULTS
In recent technology we have seen that how the advance image processing techniques with the help of pattern recognition and
artificial intelligence can be effectively used in automatic detection and classification of various facial signals. Among them face
recognition and facial expression recognition are the best ones to describe the concept of man-machine interaction efficiently. In
both of these techniques we are doing patter recognition. For example, in face recognition we consider two patterns known and
unknown whereas in facial expression recognition we consider five patterns neutral, happy, sad, angry, disgust etc.
Facial expression recognition can be used in behavior monitor system and medical system. In this paper we will show how the
concept of face recognition with the help of neural network can be used in facial expression recognition process. The following
figure shows the concept of a typical facial expression recognition system.
Figure 1: Block diagram of a typical facial expression recognition system.
At first we need to acquire the image on which we will apply our facial expression recognition techniques. Input image can be
captured by any kind of imaging system. If the Input Image is colour (RGB), then convert it to Gray scale image.
The input image can be of different size, format, colour (RGB or gray) etc. Hence we should preprocess the input image, so
that we can efficiently apply our algorithm to get better result. In the preprocessing technique we use some compression
technique like 2D-DCT to compress the data, because an image consists a large number of data, which increases the computation
time. Also, we can apply some filtering techniques to remove the noise from the input image, because the presence of any
artifacts can lead to false detection of facial features, which could produce wrong output.
As in all imaging process, artifacts can occur, resulting in degraded quality of image which can compromise imaging
evaluation. An artifact is a feature appearing in an image that is not present in the original object. Artefacts remain a problematic
area and it affects the quality of the image. Pre-processing (artefact removal) techniques are used to improve the detection of the
unwanted portion from the given images.
Algorithm for Artefact Removal:
1) Step 1. Grayscale facial images are taken as input.
2) Step 2.Threshold value of the image is calculated using the standard deviation technique described above.
3) Step 3. The image is binarized using the threshold value. i.e. pixels having value greater than the threshold is set to 1
and pixels less than the threshold are set to 0.
4) Step 4. The binarized image is labelled and areas of connected components are calculated using equivalence classes.
5) Step 5. The connected component with the maximum area and the connected component with the second highest area
are found out.
6) Step 6. The ratio of the maximum area to that of second maximum area are calculated.
33

(J4R/ Volume 02 / Issue 02 / 006)
7) Step 7. On the basis of the ratio if ratio is high only the component with highest area is kept and all others are removed
otherwise if ratio is low the component with the highest and second highest area are kept and all others are removed.
8) Step 8. A convex hull is calculated for the one pixel in the image and all regions within the convex hull are set to one.
9) Step 9.Now the above obtained image matrix is multiplied to the original image matrix to obtain an image consisting of
only medical image without any artefact.
In RGB images each pixel has a particular colour; that colour is described by the amount of red, green and blue in it. If each of
these components has a range 0255, this gives a total of 256^3 different possible colours. Such an image is a stack of three
matrices; representing the red, green and blue values for each pixel. This means that for every pixel there correspond 3 values.
Whereas in greyscale each pixel is a shade of gray, normally from 0 (black) to 255 (white). This range means that each pixel can
be represented by eight bits, or exactly one byte. Other greyscale ranges are used, but generally they are a power of 2.so,we can
say gray image takes less space in memory in comparison to RGB images.
Edge detection refers to the process of identifying and locating sharp discontinuities in an image. The discontinuities are
abrupt changes in pixel intensity which characterize boundaries of objects in a scene. Edges characterize boundaries and are
therefore a problem of fundamental importance in image processing and an important tool for image segmentation. The concept
of edge is highly useful in dealing with regions and boundaries as an edge point is transition in gray level associated with a point
with respect to its background. Edges typically occur on the boundary between two regions. The following algorithm is used for
edge detection as pre-processing step.
Algorithm for Edge Detection:
Basic functions are defined which are used in algorithms:
Parent (i)
Return i/2
Left (i)
Return 2i
Right (i)
Return 2i+1
Total number of nodes in a Complete Binary Tree
tNode (h)
Return 2h-1
The number of terminal nodes (leaf nodes) in a Complete Binary Tree
lNode (h)
Return 2h-1
The number of internal nodes (non-leaf nodes) in a Complete Binary Tree
iNode (h)
Return tNode (h) lNode (h)
Algorithms for: Storing Original Colour Space at Leaf Nodes of Tree
ORIGINAL-HISTOGRAM (Image, height, width)
Loop x 1 to height
Do Loop y 1 to width
Do Intensity Image [x, y]
Tree [(iNode (h) + 1) + Intensity].count Tree [(iNode (h) + 1) + Intensity].count + 1
xx +1
yy +1
Return Tree
Algorithms for: generate quantize colour spaces in different level of tree
LEVEL-HISTOGRAM (Tree)
Loop x iNode (h) + 1 To tNode (h)
Do Lcount Tree (x).count
Loop y Parent (x) down To 0
Do
If
x mod 2 0
Then Tree [y].intensity Tree [x].intensity
Tree [y].count Tree [y].count + Lcount
Else If Tree [y].count Tree [x].count
ThenTree [y].intensity Tree [x].intensity
Tree [y].count Tree [y].count + Lcount
xy
34

(J4R/ Volume 02 / Issue 02 / 006)
y Parent (x)
xx+1
Return Tree
Algorithms to: Calculate the Average Bin Distance
BIN-DISTANCE (Tree, h1)
TotBin 0
TotBinDist 0
Loop x iNode (h1) + 2 to tNode (h1)
Do TotBin TotBin + 1
TotBinDist TotBinDist + (Tree [x] .intensity - Tree[x - 1] .intensity)
xx+1
AvgBinDist TotBinDist / TotBin
Return AvgBinDist
Algorithms for: Calculation of MDT by Identifying the Prominent Bins and Truncate the Non-Prominent Bins
CALCULATE-MDT (Tree, h1)
Tree [iNode (h1) + 1].prominent 1
TotPrmBin 0
TotPrmBinDist 0
Loop x iNode (h1) + 2 to tNode (h1)
Do If Tree [x] .intensity - Tree [x - 1] .intensity AvgBinDist
ThenTree[x].prominent 1
TotPrmBin TotPrmBin + 1
TotPrmBinDist TotPrmBinDist + (Tree [x] .intensity - Tree [x - 1] .intensity) Else
Tree[x].prominent 0
xx+1
MDT TotPrmBinDist / TotPrmBin
Return MDT
Algorithms for: Redraw the Image Using Truncated Histogram
REDRAW - IMAGE (Image, height, width, Tree, h1, h)
Loop x 1 to height
Do NewIntensity (Image [x, y] / (tNode (h) / tNode (h1))) + 1
If Tree[iNode (h1) + NewIntensity + 1].prominent 1
Then While Tree [iNode (h1) + NewIntensity + 1].prominent 1
Do NewIntensity NewIntensity 1
NewImage [x, y] NewIntensity
yy+1
xx+1
Return NewImage
Algorithms for: derive the Horizontal Edge of the image
HozEdgeMap (NewImage, height, width, MDT)
Loop x 1 to height
Do flag 1
Loop y 1 to width
Do If Flag = 1
Then NewIntensity NewImage [x, y]
NxtNewIntensity NewImage [x, y]
If |NewIntensity NxtNewIntensity | MDT
Then Flag 1
HozEdgeMapImage [x, y] BLACK
Else
Flag 0
HozEdgeMapImage [x, y] WHITE
y y + 1
x x + 1
Return HozEdgeMapImage
35

(J4R/ Volume 02 / Issue 02 / 006)
Algorithms for: derive the Edge of the image

EDGEMAP (HozEdgeMapImage, VerEdgeMapImage, height, width)
Loop x 1 to height
Do EdgeMapImage [x, y] HozEdgeMapImage [x, y] OR VerEdgeMapImage [x,
xx+1
Return EdgeMapImage
The results obtained are shown Figure 2.
(a)
Fig. 2: (a) Template of the Face
(b)
(b) Edge Detection of Side Face
y]
yy+1
(c)
(c) Edge Detection of Front Face
Once we acquire the face image, we need to extract the facial features from the background of the image. In this paper we use
skin colour based face detection technique, which uses RGB and HSV colour model. There are another colour model (YCbCr)
that can also be used to detect skin colour region.
We use 2D-DCT to compress the extracted facial feature, which can make our processing faster. As our algorithm uses an
image database, we have to apply the compression technique in all the images in the database. An example image is shown in the
figure 5 on which we want to execute our proposed algorithm.
Fig. 3: Original image containing

a single frontal viewed face.
Fig. 4: Image in HSV colour space Fig. 5: Extracted skin colour region from
the image
Fig. 6: Binary image showing region boundary Fig. 7: input Image

The following algorithm extracts facial features.
Fig. 8: Detected Face Region
Facial_Feature_Detection (Input Image, Template Images)

Step1. Start
Step2. Read Input Human Face Image.
If the Input Image is colour (RGB), then
convert it to Gray scale Image and save the pixel values to a 2D array let gface.
Else
save the pixel values of the input image to a 2D array let gface.
Step3. Read Left eye template image.
36

(J4R/ Volume 02 / Issue 02 / 006)
If the template image is colour (RGB), then

convert it to Gray scale Image and save the pixel values to a 2D array let gleft.
Else
save the pixel values of the input image to a 2D array let gleft.
Step4. Read Right eye template image.
convert it to Gray scale Image and save the pixel values to a 2D array let gright.
Else
save the pixel values of the input image to a 2D array let gright.
Step5. Read Nose template image.
convert it to Gray scale Image and save the pixel values to a 2D array let gnose.
Else
save the pixel values of the input image to a 2D array let gnose.
Step6. Read Mouth template image.
convert it to Gray scale Image and save the pixel values to a 2D array let gmouth.
Else
save the pixel values of the input image to a 2D array let gmouth.
Step7. Declare 4 2D Array C1, C2, C3 & C4 of size m*n where m*n is the size of gface.
Step8. Calcualte C1[][] = 2D_norm_crosscorr(gleft,gface)
C2[][]= 2D_norm_crosscorr(gright,gface)
C3[][] = 2D_norm_crosscorr(gnose,gface)
C4[][] = 2D_norm_crosscorr(gmouth,gface)
Step9. Call (x11,y11,w1,h1) = Find_max(C1)
(x21,y21,w2,h2) = Find_max(C2)
where (x11,y11,w1,h1), (x21,y21,w2,h2), (x31,y31,w3,h3), (x41,y41,w4,h4) are top left pixel
coordinate, width, height of the matched rectangular area around left eye, right eye,
nose and mouth respectively.
Step10. Calculate x12 = x11 + w1 & y12 = y11 + h1
x22 = x21 + w2 & y22 = y21 + h2
x32 = x31 + w3 & y32 = y31 + h3
x42 = x41 + w4 & y42 = y41 + h4
where (x12,y12), (x22,y22), (x32,y32), (x42,y42) are bottom right pixel coordinate of the
matched rectangular area around left eye, right eye, nose and mouth respectively.
Step11. Draw Boundary Rectangle around left eye in gface with top left, top right, bottom
left and bottom right pixel coordinates as (x11,y11), (x12,y11), (x11,y12) & (x12,y12)
respectively.
Draw Boundary Rectangle around right eye in gface with top left, top right, bottom
left and bottom right pixel coordinates as (x21,y21), (x22,y21), (x21,y22) & (x22,y22)
respectively.
Draw Boundary Rectangle around nose in gface with top left, top right, bottom left
and bottom right pixel coordinates as (x31,y31), (x32,y31), (x31,y32) & (x32,y32)
respectively.
Draw Boundary Rectangle around mouth in gface with top left, top right, bottom left
and bottom right pixel coordinates as (x41,y41), (x42,y41), (x41,y42) & (x42,y42)
respectively.
Calculate middle point pixel coordinate (x1mid,y1mid) of the boundary rectangle around
Left eye as x1mid = (x11+x12)/2 and y1mid = (y11+y12)/2.
Step12. Calculate Euclidian Distance between middle point pixel coordinate (x1mid,y1mid) of
the boundary rectangle around left eye and top left pixel coordinate (x41,y41) of the
boundary rectangle around mouth as:
Dist1 = {( x1mid x41)2 + ( y1mid - y41)2} unit.
Step13. Calculate middle point pixel coordinate (x2mid,y2mid) of the boundary rectangle around
right eye as x2mid = (x21+x22)/2 and y2mid = (y21+y22)/2.
Step14. Calculate Euclidian Distance between middle point pixel coordinate (x2mid,y2mid) of
the boundary rectangle around right eye and top right pixel coordinate (x42,y41) of
the boundary rectangle around mouth as:
37

(J4R/ Volume 02 / Issue 02 / 006)
Dist2 = {( x2mid x42)2 + ( y2mid - y41)2} unit.

Step15. Write the value of Dist1 and Dist2 in a output text file for comparison.
Step16. Repeat step 1 to 15 for another same human face but with smiling facial expression.
Step17. Compare both input face images according the distances measured between eyes &
mouth. The image with larger distance is considered as Happy face or smiling face,
in general,.
Step18. Exit
2D_Norm_Crosscorr (Template Gray scale Image, Input Gray scale Image)
Step1. Start
Step2. Perform 2D Cross Correlation between Template Image and Input Image pixel values
and return 2D array C of size m*n with values of the corresponding Cross correlation
where m*n is the size of the Input Image.
Step3. End
Find_Max(C[ ][ ])
Step1. Start
Step2. Find Maximum Value of 2D Array C[ ][ ] and determine the corresponding
rectangular region where the maximum value is found.
Step3. Find top left position coordinate (x,y), width (w) and height (h) of the rectangular
region and return the values.
Step4. End
Facial_expression_recognition (Input Image, 3 Training Image Databases)
Step19. Start
Step20. Read Input Human Face Image and Store the pixel values to an array let face.
Step21. Call Processed_Face = imPreprocess(face)
Step22. Set i=1
Step23. Repeat Step 6 to 9 for every Image of Train_Neutral_Other Image Database
Step24. Read the Image from the Database and Store the pixel values to an array let t.
Step25. Call t1= imPreprocess(t)
Step26. Store t1 into image cell Train_Neutral_Other_Cell as
Train_Normal_Other_Cell(1,i)=t1
Step27. Set i=i+1
Step28. Set i=1
Step29. Repeat Step 12 to 15 for every Image of Train_Smiling_Other Image Database
Step32. Store t1 into image cell Train_Smiling_Other_Cell as
Train_Smiling_Other_Cell(1,i)=t1
Step33. Set i=i+1
Step34. Set i=1
Step35. Repeat Step 18 to 21 for every Image of Train_Angry_Sad Image Database
Step38. Store t1 into image cell Train_Angry_Sad_Cell as
Train_Angry_Sad_Cell(1,i)=t1
Step39. Set i=i+1
Step40. Create 3 2D Array of size (no_of_images * mn) for 3 Training Databases where
no_of_images refers to the total number of images in the corresponding training databases
respectively and m,n refers to the predefined size mentioned in Impreprocess function. Let
traindata1 (n1 *mn), traindata2 (n2*mn) and traindata3 (n3*mn) are 3 arrays for
Train_Neutral_Other, Train_Smiling_Other and Train_Angry_Sad Training Image
Databases respectively, with n1, n2 and n3 are number of images in the corresponding
databases.
Step41. Initialize all elements of traindata1, traindata2 and traindata3 array to 0.
Step42. Set i=1
Step43. Repeat step 26 to27 for n1 times
Step44. Set traindata1(i,:)= Train_Normal_Other_Cell(1,i)
38

(J4R/ Volume 02 / Issue 02 / 006)
Step45. Set i=i+1

Step46. Set i=1
Step48. Set traindata2(i,:)= Train_Smiling_Other_Cell(1,i)
Step49. Set i=i+1
Step50. Set i=1
Step52. Set traindata3(i,:)= Train_Angry_Sad_Cell(1,i)
Step53. Set i=i+1
Step54. Create 3 1D Arrays, namely class1, class2 and class3 of size n1,n2 and n3 respectively corresponding to
Train_Neutral_Other, Train_Smiling_Other and Train_Angry_Sad Training Image Databases respectively.
Step55. Set i=1
Step56. Repeat Step 39 to 40 for all images of Train_Neutral_Other Image Database
Step57. If ith image of Train_Neutral_Other is of Neutral expression
then set class1(i)= 1
else set class1(i)= -1
Step58. Set i=i+1
Step59. Set i=1
Step60. Repeat Step 43 to 44 for all images of Train_Smiling_Other Image Database
Step61. If ith image of Train_Smiling_Other is of Smiling expression
Step62. Set i=i+1
Step63. Set i=1
Step64. Repeat Step 47 to 48 for all images of Train_Angry_Sad Image Database
Step65. If ith image of Train_Angry_Sad is of Angry expression
Step66. Set i=i+1
Step67. Call SVMTrained1=SVM_Training(traindata1,class1)
SVMTrained2=SVM_Training(traindata2,class2)
SVMTrained3=SVM_Training(traindata3,class3)
Step68. Call result1=SVM_Classify(SVMTrained1, Processed_Face)
result2=SVM_Classify(SVMTrained2, Processed_Face)
result3=SVM_Classify(SVMTrained3, Processed_Face)
Step69. Call FinalExpression=Recognize_Expression(result1,result2,result3)
Step70. Display FinalExpression as output
Step71. Exit
Impreprocess (Image_Pixel_Array)
Step1. Start
Step2. Convert Image_Pixel_Array to its corresponding double format let
Image_Pixel_Array_Double.
Step3. If Image_Pixel_Array_Double is of format a*b*3, then
convert it to Corresponding Gray scale and save the pixel values to a 2D array let
gImage.
Else
save the pixel values of the input image to a 2D array let gImage.
Step4. Resize gImage to a Predefined size say m*n & save the pixel values to a 2D array let
gImage_Resized.
Step5. Reshape gImage_Resized Array to a 2D array of size 1*(mn) & save the pixel values to a
2D array let gImage_Reshaped.
Step6. Return the Array gImage_Reshaped.
Step7. End
SVM_Training (Training_Data, Group_Membership_Class)
Step1. Start
Step2. Train Linear Support Vector Machine with Training_Data and Group_Membership_Class
and store the value in an array let SVM1.
39

(J4R/ Volume 02 / Issue 02 / 006)
Step3. Return the array SVM1

Step4. End
SVM_Classify (SVM_Trained, Img_Array)
Step1. Start
Step2. Classify Img_Array in one of the classes with SVM_Trained and SVM Binary Classifier
and store the value in a variable let Classifier1
Step3. Return the value Classifier1
Step4. End
Recognize_Expression (Val1, Val2, Val3)
Step1. Start
Step2. If Val1=1
Set Expression=Neutral
Else if Val1= -1 and Val2=2
Set Expression=Smiling
Else if Val1= -1 and Val2= -2 and Val3=3
Set Expression=Angry
Else if Val1= -1 and Val2= -2 and Val3= -3
Set Expression=Sad
Step3. Return the value of Expression
Step4. End
The following figures describes the output of the proposed system
Figure 9 Test image
Fig. 12: Detected Face Region
Figure 10 Templates of this image
Fig. 13: Neutral face
Fig. 11 Left eye template
Fig. 14: Neutral face

template matching
Fig. 15: Smiling face
IV. CONCLUSIONS
Facial expression recognition or emotion detection system has numerous applications in image processing domains, security
applications domain or any type of biometric system.
REFERENCES
[1]
[2]
Farah Azirar, Facial Expression Recognition, Bachelor Thesis, School of Electronics and Physical Sciences, Department of Electronic Engineering,
University of Surrey, 2004.
L. Franco and A. Treves. A Neural Network Face Expression Recognition System using Unsupervised Local Processing. Proceedings of the Second
International Symposium on Image and Signal Processing and Analysis (ISPA 01), Croatia, pp. 628-632, June 2001.
40

(J4R/ Volume 02 / Issue 02 / 006)
Angel Noe Martinez-Gonzalez and Victor Ayala-Ramirez, Real Time Face Detection Using Neural Networks, 2011 10th Mexican International
Conference on Artificial Intelligence.
[4] M. Suwa, N. Sugie, and K. Fujimora. A preliminary note on pattern recognition of human emotional expression. In International Joint Conference on
Pattern Recognition, pages 408410, 1978.
[5] H.A. Rowley, S. Baluja and T.Kanade Rotation Invariant Neural Network-Based Face Detection, Proc IEEE Conf. Computer Vision and Pattern
Recognition, 1998, pp 38-44.
[6] C.H. Lee, J.S. Kim, K.H. Park, Automatic Human Face Location in a Complex Background Using Motion and Colour Information, Pattern Recognition,
vol. 29, no. 11, 1996, pp. 129-140.
[7] K. Sobottka, I. Pitas, A Novel Method for Automatic Face Segmentation, Facial Feature Extraction and Tracking, Signal Process. Image Communication,
vol. 12, no. 3, 1998, 263-281.
[8] Hsu, Rein-Lien, Abdel-Mottaleb, Mohamed, Jain, and Anil K. Face Detection in Colour Images. IEEE Trans. Pattern Analysis and Machine Intelligence,
vol. 24, no. 5, 2002, pp. 696-706.
[9] A. Pentland and M. Turk. Eigen faces for recognition. Journal of Cognitive Neuroscience, 3(1):71-86, 1991.
[10] H. A. Rowley. Neural Network-Based Face Detection. PhD thesis, Carnegie Mellon University, Pitsburgh, 1999.
[3]
41

Detecting Facial Expression in Images

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Detecting Facial Expression in Images

Hochgeladen von

Copyright:

Verfügbare Formate

Journal for Research | Volume 02 | Issue 02 | April 2016