Download Facial Expression Recognition-Implementation and Applications In Computer Sciences-Project Presentation and more Slides Applications of Computer Sciences in PDF only on Docsity!
Facial Expression Recognition: An Overview
Definition:
Facial expression recognition deals with the
classification of facial features into classes based
on visual information.
Guessing the meaning of facial deformations
Three phases of recognition:
Face Acquisition
Facial Feature Extraction
Expression Classification
Facial Expression Analysis framework
Most of the facial expression recognition systems
make use a simple facial expression analysis
framework
Facial Expression Recognition Techniques
Feature Extraction Techniques:
PCA
Wavelets
Feature Point Tracking
Optical Flow
Expression Classification Techniques:
Learning Vector Quantization (LVQ).
Support Vector Machines (SVM)
Hidden Markov Models (HMM)
Feed-Forward Back Propagation Network
Facial Expression Recognition Using PCA
PCA
Principle Component Analysis
A technique useful for compression and
classification of data
Purpose
To reduce the dimensionality of the dataset
Finding new set of variables, smaller than the
original set of variables, that contain most of the
sample’s information.
Facial Expression Recognition Using PCA
Step 2: Feature Extraction
Calculate the Mean of the Matrix X.
Subtract Mean from All Vectors in X. i.e.
A = X - Mean(X)
Get the covariance matrix
the covariance matrix is a matrix of covariance between
elements of a vector.
L = A'*A;
Get the eigenvectors and eigenvalues
[Vectors,Values] = eig(L);
Facial Expression Recognition Using PCA
Step 3: Expression Classification
For classification i used LVQ.
Create a learning vector quantization network.
Goal is to have the network "discover" structure in the
data by finding how the data is clustered.
In Matlab, an LVQ network can be created with the
function newlvq
We require input vectors and target expression classes
to train the neural network.
Training Results
Training Results
No. of Misclassificati ons
36/
Recognition Rate 80.15%
Error Rate 18.85%
Confusion Matrix of Training Result
Anger Disgust Fear Happy Natural Sad Surprise
Anger 25/ =92.6%
0% 0% 0% 0% 2/27 =7.41% 0%
Disgust 4/ =15.4%
19/ =73.08%
2/ =7.69%
0% 1/ =3.85%
0% 0%
Fear 0% 0% 22/ =78.57%
0% 2/ =7.14%
4/ =14.29%
0%
Happy 1/ =3.57%
0% 1/ =3.57%
25/ =89.29%
1/ =3.57%
0% 0%
Natural 0% 0% 1/ =3.70%
3/ =11.11%
18/ =66.67%
5/ =18.52%
0%
Sad 4/ =14.3%
0% 0% 1/ =3.57%
2/ =7.14%
21/ =75%
0%
Surprise 0% 0% 1/ =3.7%
0% 0% 0% 26/ =96.3%
Result Improvement
Two methods are used
By changing the range and number of Principle
Components
By Extracting Mouth and Eyes and then apply
Principle Component Analysis
Result Improvement I
PCA is also used for face recognition
applications.
The Principle components that have higher
Eigen values may or may not correspond to
varying expression.
i have used first 50 Principle Components
But in order to improve the results i have tried
different ranges of principle components
Result Improvement Using 20-
Principle Components
Recognition Using 20-50 PC's and 500 Epochs
No. of Misclassifications
10/
Recognition Rate 95.31%
Error Rate 4.69%
Result Improvement II
For improvement, following steps are used
Extract Both eyes from image
Extract mouth from image
Reshape images into vectors
Combine all vectors in form of Matrix
Apply PCA on Matrix for Feature extraction
Apply LVQ for Feature Classification
Training Results
Recognition Using 20- PC's and 150 Epochs
No. of Misclassificatio ns
11/
Recognition Rate
94.24%
Error Rate 5.76%
Testing Results
Testing Results
No. of Misclassifications 10/
Recognition Rate 54.54%
Error Rate 45.46%
