Human emotion recognition

11
A Statistical Shape and Texture- based Technique for Recognizing Facial Emotions Anukriti Dureha 7CSE2 A2305210153

Transcript of Human emotion recognition

Page 1: Human emotion recognition

A Statistical Shape and Texture- based Technique for Recognizing Facial Emotions

Anukriti Dureha7CSE2A2305210153

Page 2: Human emotion recognition

Project Objective

To be able to recognize one of the seven universal emotions from a facial image based on different facial features.

Neutral

Joy

Sadness

Surprise

Fear

Anger

Disgust

Seven Universal Expressions

Page 3: Human emotion recognition

Project Methodology

Page 4: Human emotion recognition

Labeling

Landmark Features

Eye Brows

Nose

LipsChin

Eyes

No. of training images 70

Total no. of landmark points

72

Face profile Frontal Faces

Datasets used Extended Cohn- Kanade

Example of a hand–labeled

image

Page 5: Human emotion recognition

Shape Modeling

Page 6: Human emotion recognition

Shape Alignment

A self-Written Module

Page 7: Human emotion recognition

Principle Component Analysis

A self-Written Module

Page 8: Human emotion recognition

Model Fitting-ASM Search

• Initialize shape parameters to zero.•Generate the model point positions.• Find pose parameters using

procrustes analysis • Projection of target image points

into model frame•Update shape parameters.• Iterate until convergence.

Page 9: Human emotion recognition

Results-Alignment of Shapes

Original Set of Images

Aligned Set of

Images

Page 10: Human emotion recognition

Results-Mean Models

Joy

AngryFear

Surprise Sad

Neutral

Page 11: Human emotion recognition

THANK YOU!