Facial expression identification using four bit co- occurrence matrixfeatures and k-nn classifier

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 04 Issue: 12 | Dec-2015, Available @ http://www.ijret.org 81
FACIAL EXPRESSION IDENTIFICATION USING FOUR-BIT CO-
OCCURRENCE MATRIXFEATURES AND K-NN CLASSIFIER
Bonagiri C S K Sunil Kumar1
, V Bala Shankar2
, Pullela S V V S R Kumar3
1,2,3
Department of Computer Science & Engineering, Aditya College of Engineering, Surampalem, East Godavari
District, Andhra Pradesh, India
Abstract
The present paper proposes a novel approach for facial expression identification based on the statistical features extracted from
the 4-bit co-occurrence matrix. In the present, first color image is converted into grey level image using 7-bit quantization
technique. To reduce the complexity of the Grey Level Co-occurrence matrix the present paper generates the co-occurrence
matrix on 4 bit grey image which is generated by quantization technique. From The 4-bit CM extract 4 features; contrast,
homogeneity, energy and correlation and generate the feature vector. By using the K-NN Classification technique classify the test
image and extract the four statistical features and identify the type of expression of the test facial image. The present method got
better comparative results when it is tested on Japanese Female Facial Expression (JAFFE) Database.
Key Words: Facial Expression identification, Classification, k-NN classifier, Quantization, and GLCM
--------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
One of the rapidly growing fields of research is facial
expression recognition due to continually growing interest in
applications for automatic human behavior psychoanalysis
and new technologies for human-machine announcement
and multimedia retrieval. Mostly systems for Facial
Expression Recognition and its Analysis proposed in the
literature focused on detecting the occurrence of
expressions, often using basic emotions or the Facial Action
Coding System (FACS Action Units, AUs,[1]. In reality,
expressions vary greatly in concentration, and concentration
is often a strong cue for the explanation of the sense of
expressions.
So many approaches are available in literature for the
recognition of the facial expressions [2, 3]. The proposed
method included the model based techniques for analyzing
the face, which is considered as one global pattern without
decomposing the face to different components. A feature
vector, representing the expression information, is obtained
by preprocessing the face image and applying different
transformations for facial expression feature extraction, like
Gabor wavelet [4], curvelets [5, 6], and local binary pattern
[7, 8, 18, 19]. Principal component analysis (PCA) is mainly
used for dimensionality reduction, and multilayer neural
networks or support vector machines are widely used for
image classification [9, 10]. The analytic or geometric
feature based methods are used to divide the face into
smaller components or subsection using which we can
identify the expressions. Characteristic points are detected,
and distances between Characteristic points are calculated to
form the feature vectors or build a model of appearance [11,
12, 13] and support vector machines SVM [14].
The above approached methods have its own disadvantages
and own demerits especially in the form of computational
time and percentage of recognition. To overcome such
disadvantages, the present paper derives a concept for the
facial expression based on the statistical features derived
from 16 gray values co-occurrence matrix. The paper is
organized into 4 sections. Section 2 describe the proposed
method and section 3 describes the results and discussions
and conclusions are given in section 4.
2. METHODOLOGY: FACIAL EXPRESSION
RECOGNITION SYSTEM
The proposed method mainly consists of 5 steps. In the first
step, crop the given input image which covers the all facial
skin of the image. In the second step, convert the cropped
color image into 4-bit quantization technique. In the third
step, generate the co-occurrence matrix which consists of 16
shades. Extract the 4 statistical features and generate the
Feature Vector in the forth step. Based on the FV values and
k-NN classifier recognize the type of expression in the last
step. The block diagram of the facial expression recognition
is shown in figure 1.
Color
Image
Cropped
Image
4-bit
quantization
k-NN
Classifier
16 shades
Grey level
Image
Generate the
Co-occurrence
Matrix
Extract the
4 statistical
features
Identify the
Facial
Expression
Fig -1: Block Diagram of the Facial expression
Identification
Step 1 : Crop the given input image which covers the all
facial skin of the image
Step 2: Crop the grey scale image: The gray facial image is
cropped to cover whole skin area of the facial image. For

_______________________________________________________________________________________
efficient identification of the expression, cropping is
necessary. The cropped image consists of forehead; two-
eyes nose, mount, and chin area. All these appearance
changed according the type of expression. Figure 2 shows
an example of the original facial image and the cropped
image.
(a) (b)
Fig -2: (a) original image (b) cropped image
Step 3: Convert the Color Image into 4-bit gray scale image
using 4-bit quantization technique
The size of the Grey Level Concurrence Matrix (GLCM)
depends on number of shades in the image. Generally, Gray
Scale image has 256 shades and the size of the GLCM
matrix is 256. So, in order to reduce the complexity of
GLCM matrix, the present paper uses the 16 shades to
generate the co occurrence matrix. 16 shades are represented
by using 4 –bits. It can be accomplished by using
quantization technique. The present paper uses 4-bit Binary
Code quantization technique.
In 4-bit Binary Code quantization technique, the original
images are quantized into 16shades of gray image is
computed from the 256 gray shades images and then the
statistical information of facial image is calculated to
describe features of the image
To extract gray level features from color information, the
present paper utilized the Gray shades which quantize the
gray shades into 4-bins to obtain 16 gray levels. The index
matrix of 16values gray image is denoted as GS(x, y). The
Gray quantization process is done by using 4-bit binary code
of 16 colors as follows:
GS(x,y)= 4*I(Red)+2*I(Green)+I(Blue) where
I(Red)= 0, 0≤Red≤16, I(Red)=k,
((4*k)+1) ≤ Red ≤ (4*(k+1)) k = [1, 2, 3, 4]
I(Green)= 0, 0≤Green≤16, I(Green)= k,
((4*k)+1) ≤ Green ≤ (4*(k+1)) k = [1, 2, 3, 4]
I(Blue)= 0, 0≤Blue≤16, I(Blur)= k,
((4*k)+1) ≤ Blue ≤ (4*(k+1)) k = [1, 2, 3, 4]
Therefore, each value of GS(x, y) is a 4 bit binary code
ranging from 0 to 15. The process of 4-bit gray scale image
generation is depicted in figure 3
139 128 131 134 132 9 8 8 8 8
142 140 127 133 138 9 9 8 8 9
142 124 129 145 142 9 8 8 9 9
141 129 128 131 133 9 8 8 8 8
137 134 134 137 137 9 8 8 9 9
(a) (b)
Fig -3: 4-bit quantization procedure a) Sample Gray level
Image b) Generated 4-bit 16 shade gray image
Step 4: generate the 4 bit grey scale concurrence matrix:
Grey level co-occurrence matrix (GLCM) is the one of the
simplest technique for describing texture image[15]. The
GLCM is gives complete and detailed information about the
image. Generally the size of co-occurrence matrix of image
number of gray level on the image. For Generating GLCM
of the Gray level image, the size of the GLCM is 256×256.
This I is the one of the drawback of GLCM. It takes more
computational time for generating the co-occurrence matrix.
To avoid such problem, the present paper generates the co-
occurrence matrix on 16 shades image so that the size of the
GLCM is reduced to 16×16 and its causes reduces the
computational cost drastically. The generated matrix called
Four-bit Co-occurrence Matrix (FCoM).
A set of statistical features defined by Haralick [15] are
extracted on FCoM of the facial image. The features used in
this method are energy, contrast, homogeneity, and
correlation. The equations of the considered feature are
shown in equations 1 to 4. The proposed FCoM combines
the merits of both statistical computation cost for processing
of the facial images. The GLCM gives the whole
information of the facial expression image
3. RESULTS AND DISCUSSIONS
To prove the functionality of the proposed scheme, it has
established a database contains 213 facial expression images
of females from Japanese Female Facial Expression
(JAFFE) Database. The Database was collected by Kamachi
and Gyoba at Kyushu University[16], Japan. Original

_______________________________________________________________________________________
images have been rescaled and cropped such that the eyes
are roughly at the same position within a distance of 60
pixels in the final images (resolution: 256 × 256 pixels). The
images include the seven categories of expressions (neutral,
sadness, surprise, happiness, anger, disgust and fear) is
roughly the same. A few of them are shown in Figure. 4.
Fig -4: Facial expression database: Examples
In the proposed method the test images are grouped into
seven categories based on expression (neutral, sadness,
surprise, happiness, disgust, anger, and fear). The four
statistical features are extracted from JAFFE database and
generate the feature vector. FCoM of seven category results
are stored in the feature vector. Feature vector directs to
demonstration of features of the FAFFE images. The
statistical features of seven groups of facial expressions are
shown in tables1, 2, 3, 4, 5, 6, and 7 respectively. To prove
the efficiency of the proposed method, took100 different
facial expressions from JAFFE data base, Google database
and scanned images and extracted the statistical features
from those images and the results are stored in the test
vector. Estimate the minimum distance between feature
vector and test vector and stored the result in the library.
Classify the test images into appropriate expression category
using minimum distance K-nearest neighbor classifier.
Some of the successful classification results of test data
based on the proposed method are shown in table 8. The
classification results of the proposed method are shown in
table 9 and corresponding classification graph is shown in
figure 5.
Table 1: Feature set values of Anger expression images
S.
No
Image
Name
Correla-
tion
Contrast Energy Homogeneity
1 Img.104 -0.0358 32.33 0.028 0.2128
2 Img.105 -0.0317 31.39 0.028 0.2117
3 Img.106 -0.0683 31.67 0.033 0.2097
4 Img.125 0.0313 30.27 0.035 0.2404
5 Img.127 0.0499 30.75 0.032 0.2439
6 Img.146 0.0306 30.14 0.032 0.2407
7 Img.147 0.0387 29.93 0.027 0.2315
8 Img.148 0.0496 29.47 0.027 0.2280
9 Img.167 0.0013 33.94 0.037 0.2421
10 Img.168 0.0120 32.72 0.035 0.2350
11 Img.169 0.0117 32.61 0.032 0.2375
12 Img.17 0.0518 29.65 0.038 0.2616
13 Img.18 0.0568 30.45 0.036 0.2459
14 Img.19 0.0621 30.19 0.043 0.2627
15 Img.190 0.0595 29.75 0.033 0.2477
16 Img.191 0.0577 30.92 0.035 0.2479
17 Img.192 0.0484 30.58 0.034 0.2439
18 Img.211 0.0175 32.86 0.037 0.2344
19 Img.212 0.0560 31.45 0.037 0.2503
20 Img.213 0.0582 30.83 0.032 0.2492

_______________________________________________________________________________________
Table 2: Feature set values of Disgust expression images
S
No
Image
Name Correlation Contrast Homogeneity Energy
1 Img 1.193 0.086 28.150 0.273 0.033
2 Img 86 -0.183 34.870 0.179 0.039
3 Img.107 -0.057 30.060 0.198 0.036
4 Img.128 0.032 29.390 0.235 0.037
5 Img.129 0.031 29.560 0.240 0.037
6 Img.130 0.047 30.800 0.244 0.038
7 Img.149 0.015 28.173 0.225 0.036
8 Img.150 0.096 26.264 0.242 0.041
9 Img.151 0.078 27.842 0.249 0.039
10 Img.170 0.015 31.270 0.245 0.038
11 Img.171 0.002 32.820 0.231 0.043
12 Img.172 -0.037 32.520 0.221 0.043
13 Img.173 0.032 32.030 0.244 0.040
14 Img.194 0.063 27.168 0.257 0.036
15 Img.195 0.018 29.784 0.238 0.040
16 Img.20 0.018 30.520 0.245 0.044
17 Img.215 0.067 30.150 0.240 0.040
18 Img.216 0.045 30.400 0.244 0.039
19 Img.22 0.082 27.590 0.255 0.044
20 Img.42 0.038 29.449 0.245 0.036
Table 3: Feature set values of Disgust expression images
Sno
Image
1 img 117 -0.219 29.82 0.216 0.033
2 img 118 -0.247 30.83 0.216 0.034
3 img 137 -0.154 28.36 0.242 0.034
4 img 138 -0.124 28.09 0.247 0.036
5 img 139 -0.192 29.73 0.236 0.032
6 img 202 -0.202 31.43 0.236 0.035
7 img 203 -0.23 30.89 0.23 0.034
8 img 204 -0.28 29.6 0.201 0.031
9 img132 -0.167 31.17 0.245 0.038
10 img152 -0.089 27.78 0.277 0.042
11 img153 -0.138 29.84 0.251 0.041
12 img159 -0.21 31.87 0.234 0.038
13 img160 -0.21 31.79 0.23 0.035
14 img180 -0.197 30.51 0.24 0.035
15 img181 -0.211 30.67 0.226 0.031
16 img182 -0.149 29.69 0.259 0.036
17 img21 -0.219 31.14 0.211 0.031
18 img30 -0.203 30.71 0.218 0.034
19 img31 -0.222 31.19 0.217 0.031
Table 4: Feature set values of Netural expression images
Sno
Image
1 img 1 0.0514 30.76 0.254 0.044
2 img 113 -0.0336 30.78 0.219 0.039
3 img 114 0.0034 30.24 0.222 0.037
4 img 115 -0.0436 30.68 0.212 0.035
5 img 134 0.0884 27.55 0.264 0.045
6 img 135 0.1134 27.34 0.272 0.046
7 img 136 0.1104 26.24 0.266 0.047
8 img 155 0.0114 30.7 0.248 0.045
9 img 156 0.0344 31.71 0.24 0.042
10 img 157 -0.0036 30.4 0.224 0.033
11 img 199 -0.0636 29.86 0.202 0.033
12 img 2 0.0784 29.39 0.25 0.04
13 img 200 -0.0216 31.22 0.222 0.035
14 img 201 -0.0126 28.78 0.212 0.032
15 img 26 0.0464 30.56 0.236 0.037
16 img 27 -0.0146 31.78 0.222 0.037
17 img 28 -0.0366 31.83 0.214 0.038
18 img 3 0.1184 28.13 0.268 0.047
19 img 49 0.1374 28.26 0.288 0.054
20 img 50 0.1354 27.34 0.295 0.057
Table 5: feature set values of Sadness expression images
Sno
Image
Name Contrast Correlation Energy Homogeneity
1 img 10 29.14 0.0585 0.046 0.259
2 img 100 30.6 -0.0445 0.033 0.203
3 img 11 27.9 0.0075 0.035 0.224
4 img 119 30.38 0.0155 0.041 0.228
5 img 12 28.76 0.0035 0.039 0.236
6 img 120 29.8 -0.0605 0.036 0.212
7 img 121 30.54 -0.0005 0.039 0.224
8 img 140 27.19 0.0975 0.036 0.248
9 img 141 28.01 0.0585 0.033 0.232
10 img 142 28.67 0.0395 0.039 0.237
11 img 161 31.21 0.0295 0.043 0.248
12 img 162 31.34 0.0335 0.043 0.245
13 img 163 29.55 0.0615 0.041 0.251
14 img 184 28.99 -0.0195 0.036 0.232
15 img 185 29.86 0.0505 0.04 0.248
16 img 186 29.74 0.0605 0.041 0.255
17 img 205 30.02 0.0365 0.038 0.236
18 img 206 29.53 0.0705 0.037 0.251
19 img 207 30.82 -0.0185 0.034 0.218
20 img 33 31.12 -0.0195 0.037 0.23

_______________________________________________________________________________________
Table 6: Feature set values of fear expression images
Sno
Image
1 img 174 -0.003 31.78 0.242 0.042
2 img 175 0.056 30.38 0.247 0.044
3 img 176 -0.023 32.2 0.242 0.044
4 img110 -0.057 30.91 0.213 0.036
5 img111 -0.07 30.81 0.205 0.033
6 img112 -0.079 31.92 0.207 0.038
7 img131 0.067 29.54 0.243 0.041
8 img132 0.02 30.49 0.231 0.038
9 img133 0.022 31.52 0.244 0.044
10 img152 0.044 28.75 0.253 0.041
11 img153 0.088 28.32 0.253 0.039
12 img154 0.089 27.65 0.255 0.041
13 img196 0.1 29.03 0.262 0.041
14 img197 0.121 28.37 0.272 0.045
15 img198 0.058 30.41 0.263 0.047
16 img217 -0.025 32.01 0.223 0.038
17 img218 0.006 31.04 0.23 0.037
18 img219 0.03 29.77 0.244 0.036
19 img23 0.064 30.6 0.262 0.046
20 img24 0.013 28.32 0.227 0.038
Table 7: Feature set values of Surprise expression images
Sno
Image
1 img101 -0.074 31.7 0.2 0.034
2 img102 -0.083 31.63 0.194 0.032
3 img103 -0.044 30.65 0.206 0.034
4 img122 -0.049 30.96 0.21 0.036
5 img123 0.01 30.69 0.229 0.038
6 img124 -0.031 30.49 0.222 0.036
7 img14 0.048 30.51 0.246 0.043
8 img143 0.055 29.27 0.243 0.038
9 img144 0.058 28.02 0.246 0.04
10 img145 0.039 29.07 0.234 0.035
11 img15 0.037 30.39 0.253 0.045
12 img16 0.063 29.42 0.253 0.042
13 img164 0.037 30.03 0.239 0.038
14 img165 0.019 31.08 0.232 0.04
15 img166 0.047 31.01 0.236 0.041
16 img187 0.032 29.05 0.244 0.039
17 img188 0.046 28.84 0.249 0.04
18 img189 0.041 29.58 0.239 0.039
19 img208 -0.03 32.08 0.22 0.038
20 img209 0.045 30.14 0.236 0.037
Table 8: Feature values of Test database images
Sno Image Name Correlation Contrast Homogeneity Energy Group Result
1 ing190 -0.11366 34.4 0.198 0.033 Anger correct
2 img144 -0.03 32.08 0.22 0.038 Surprise correct
3 img156 -0.0126 28.78 0.212 0.032 Neutral correct
4 gog_im1 -0.074 31.7 0.2 0.034 Surprise correct
5 gog_im2 0.0395 28.67 0.237 0.039 Neutral correct
6 gog_im3 -0.0195 28.99 0.232 0.036 Sadness correct
7 gog_im4 0.063164 29.32 0.248 0.036 Disgust correct
8 gog_im5 -0.179 28.82 0.221 0.028 Happiness correct
9 si-08 -0.003 31.78 0.242 0.042 Fear correct
10 si-09 0.066517 29.99 0.24 0.04 Disgust correct
11 si-10 -0.21 31.79 0.23 0.035 Happiness correct
12 si-11 -0.023 32.2 0.242 0.044 Fear correct
13 img140 0.0035 28.76 0.236 0.039 Sadness correct
14 img65 0.001981 32.28 0.23 0.043 Anger correct
15 img204 -0.247 30.83 0.216 0.034 Happiness correct
16 img131 0.06 28.65 0.261 0.045 Fear correct
17 gog_im11 0.045202 30.042 0.244 0.039 Disgust correct
18 gog_im12 -0.209 30.19 0.215 0.032 Happiness correct
19 si-01 0.0784 29.39 0.25 0.04 Neutral correct
20 gog_im6 -0.003 31.78 0.242 0.042 Fear correct
21 gog_im7 -0.0486 30.59 0.209 0.035 Neutral correct
22 gog_im8 -0.03165 31.93 0.211 0.028 Anger correct
23 gog_im9 0.037 30.03 0.239 0.038 Surprise correct
24 gog_im10 0.0344 31.71 0.24 0.042 Neutral correct
25 si-02 0.049949 30.57 0.243 0.032 Anger correct

_______________________________________________________________________________________
Table 9: Classification Results of the test database
Data
Base
Name
No of
Images
Correctly
Classified
Not
correctly
classified
%
recognition
JAFFE 60 59 1 98.33
Google 20 19 1 95.00
Scanned 20 19 1 95.00
3. CONCLUSIONS
The present paper proposed a new scheme for generating the
gray scale image with 16 shades. The novelty of the present
approach is that its take less time to generate the co-
occurrence matrix because of the image has only 16 gray
levels. The proposed method has got the %recognition is
about 96.11 when applied on 3 different databases.
Fig -5: Classification graph of the proposed method
REFERENCES
[1] M. Bartlett, G. Littlewort, M. Frank, C. Lainscsek, I.
Fasel, and J. Movellan. Automatic recognition of facial
actions in spontaneous expressions. Journal of
Multimedia, 1(6):22–35, 2006.
[2] B. Fasel and J. Luettin. Automatic facial expression
analysis: a survey.Pattern Recognition, 36(1):259–275,
2003.
[3] Z. Zeng, M. Pantic, G. I. Roisman, and T. S. Huang. A
survey of affectrecognition methods: audio, visual, and
spontaneous expressions.IEEE Trans. Pattern Analysis
nd Machine Intelligence, 31(1):39–58,2009
[4] Juliano J. Bazzo Marcus V. Lamar, “Recognizing
Facial Actions Using Gabor Wavelets with Neutral
Face Average Difference,” Sixth IEEE International
Conference on Automatic Face and Gesture
Recognition (FGR’04), 2004.
[5] Xianxing Wu and Jieyu Zhao, "Curvelet Feature
Extraction for Face Recognition and Facial Expression
Recognition", Sixth International
[6] Conference on Natural Computation, pp.1212-1216,
2010
[7] E. Candés, L. Demanet, D. Donoho and L. Ying, “Fast
Discrete Curvelet Transforms,”Applied and
Computational Mathematics, Caltech, Pasadena, 2006.
[8] S. Singh and R. Maurya, A. Mittal,"Application of
Complete Local Binary Pattern Method for Facial
Expression Recognition",IEEE Proceedings of 4th
International Conference on Intelligent Human
Computer Interaction, Kharagpur, India, December 27-
29, 2012.
[9] R.Hablani,N. Chaudhari andS.Tanwani "Recognition of
Facial Expressions using Local Binary Patternsof
Important Facial Parts," International Journal of Image
Processing (IJIP), vol. 7 (2), 2013.
[10]M.Pantic and L.J.M. Rothkrantz, “Expert System For
Automatic Analysis Of Facial Expressions,” Image And
Vision Computing,vol.18, pp. 881-905,2000.
[11]D. Arumugan, S. Purushothaman.”Emotion
Classification Using Facial Expression,” International
Journal of advanced Computer Science and
Application, vol. 2, no 7,2011.
[12]T. F. Cootes, G. J. Edwards and C. J. Taylor. “Active
Appearance Models,” IEEE Transaction on Pattern
Analysis and Machine Intelligence , vol. 23, n° 6, 2001.
[13]M. S. Bartlett, J. C. Hager, P. Ekman, and T. J.
Sejnowski.“Measuring facial expressions by computer
image analysis,” Psychophysiology, Cambridge
University, USA, vol. 36, pp. 253–263, 1999.
[14]Zheng You. “Feature-Based Facial Expression
Recognition: Sensitivity Analysis and Experiments with
a Multi-Layer Perceptron,” International Journal of
pattern Recognition and artificial Intelligence vol.
13(6), pp. 893-911, 1999.
[15]M. S. Bartlett, J. C. Hager, P. Ekman, and T. J.
Sejnowski.“Measuring facial expressions by computer
image analysis, ”Psychophysiology, Cambridge
University, USA, vol. 36, pp. 253–263, 1999.
[16]Robert M. Haralick, K Shanmugam and Its’Hak
Dinstein(1979). “Textural Features for Image
Classification.” IEEE Transactions on Systems, Man,
and Cybernetics pp. 610-621.
[17]M. Lyons, S. Akamatsu, M. Kamachi, and J. Gyoba.
Coding facial expressions with gabor wavelets. In
Proceedings of the Third IEEE International
Conference on Automatic Face and Gesture
Recognition, Nara, Japan, Apr. 1998
[18]SVVSR Kumar, Pullela., V.Vijaya Kumar, and
Rampay. Venkatarao. "Age Classification Based On
Integrated Approach", International Journal of Image
Graphics and Signal Processing, 2014
[19]Satyanarayana Murty, Gorti, J Sasi Kiran, and V.Vijaya
Kumar. "Facial Expression Recognition Based on
Features Derived From the Distinct LBP and GLCM",
International Journal of Image Graphics and Signal
Processing, 2014.
[20]Pullela S V V S R Kumar et.al., Classification of Facial
Expressions based on Transitions Derived from Third
Order Neighborhood LBP, Global Journal of Computer
Science and Technology, Graphics & Vision, Vol. 14
no. 1, pp. 1-12, 2014.
BIOGRAPHIES
Bonagiri C S K Sunil Kumar is pursuing
his masters degree from Aditya College of
Engineering, Surampalem. His research
interests include Data Mining and Image
processing.

_______________________________________________________________________________________
V Bala Shankar is working as Senior
Assistant Professor in the Department of
Computer Science & Engineering, Aditya
College of Engineering, Surampalem. His
research interests include Image Processing,
Cloud Computing.
Dr. Pullela S V V S R Kumar is working
as Professor in CSE at Aditya College of
Engineering, Surampalem. He received his
Ph.D degree from Acharya Nagarjuna
University. He is having more than 15 years
of experience and published 13 research
papers in various International Journals and
Conferences. He acted as a reviewer and PC Member for
various international conferences. His research interests
include Data Mining, Pattern Recognition and Image
Processing.

Facial expression identification using four bit co- occurrence matrixfeatures and k-nn classifier

More Related Content

What's hot (20)

Similar to Facial expression identification using four bit co- occurrence matrixfeatures and k-nn classifier (20)

More from eSAT Journals (20)

Recently uploaded (20)

Facial expression identification using four bit co- occurrence matrixfeatures and k-nn classifier