(Roll no- 91/MCA/090021)
Project under the guidance of
DR. AMLAN CHAKRABARTI
1
What is Emotion?
 Emotions are reflected in voice, hand and body gestures,
and mainly through facial expressions.
2
Facial Emotion
 There are six types of facial emotions.
Anger Fear Disgust Happy Sad Surprise
3
Importance of emotion recognition
 Human beings express emotions in day to day
interactions.
 Understanding emotions and knowing how to react to
people’s expressions greatly enriches the interaction.
4
Review Of Previous Work
 Image pre-processing
 Facial feature extraction
-select some point on feature region
-consider a matrix surrounding that point
-calculate distances
 Facial expression detection
Reff.: [“Optical flow based analyses to detect emotion from human facial image data”- Axel Besinger ,
Tamara Sztynda, Sara Lal , Carmen Duthoit, Johnson Agbinya, Budi Jap ,David Eager, Gamini
Dissanayake]
5
Proposed Methodology
 Capture Image
 Image Pre-processing
-RGB to gray scale conversion
-Scale Normalization
 Crop Feature Regions
 Edge Detection
-Prewitt edge detection
 Facial Emotion Classification
6
Methodology(Cont…)
7
Capture Image
 Digital camera, web-cam
 Resolution must be fixed
 Distance between the web-cam and the face has to be
fixed.
 Position of face must be fixed
 Must not wear glasses or have facial hair.
8
Capture Image(Cont…)
Normal Facial Image Emotional Facial Image
9
RGB to gray scale conversion
Color Facial Image Grayscale Facial Image
 Reduce complexity.
10
Scale Normalization
 Select only the face portion within a rectangular
frame
 Remove unnecessary information
 By defining region.
11
Crop Feature Regions
 Region of Interest: two Eyes, two Eye brows, Lip
 By defining the particular regions.
12
Edge detection
ROI ROI after edge detection
13
Algorithm:
Input: A Sample Image
Output: Detected Edges
Step 1: Accept the input image
Step 2: Apply mask h1 and h3 (prewitt mask) to the input
image.
Step 3: Masks manipulation of h1 and h3 separately on
the input image
Edge Detection (Cont…)
14
Edge detection(Cont…)
Step 4: Results combined to find the absolute magnitude of the
gradient
Where
Step 5: The absolute magnitude is the output edges.
15
Facial Emotion Classification
 Calculate displacement of the points.
 Calculate averages of those values.
 Calculate standard deviation (SD) of those averages.
 Compare this SD with a threshold.
 Get Emotion.
16
Algorithm
 Step 1: Take a still image of a normal expression pic1
(say) of a human face.
 Step 2: Converts the color image to grayscale.
 Step 3: Crop the five facial image region of interest (ROI)
(eyes, eye brows and lip) from the image by defining
region.
 Step 4: Find edges of all image region.
 Step 5: Take a still image of a emotional face (angry or
happy) pic2 (say) of same person and repeat step 2, 3
and 4.
17
Algorithm(Cont…)
 Step 7: Comparing the deviation of edges of the
specified region of pic1 with pic2 by finding the
Euclidian distances of coordinate of each pixel.
 Step 8: Put the Euclidian distances in a array k (say).
 Step 9: Find the standard deviation (SD) from the
elements of array.
 Step 10: Comparing the SD with pre-define threshold
and get the emotions.
18
Experiment & Result
 30 images were used to create the template.
 60 were tested.
Statistical Values of Three categories

19
Experiment & Result
 Total 83.33% classify Facial Emotion.
63.33% were correct
20
Experiment & Result
 80% happy Emotion classification were correct.
 50% Angry Emotion classification were correct.
 Almost 100% null classification were correct.
21
Applications
 E-Learning system
 Robotics System
 Human-Computer interaction System
22
Conclusion and Future Work
 Implementation through this process is quite easy.
 Have to improve over the capturing Process.
 ROI extracted manually
-That should be automatically.
 Edge detection procedure should be less complex.
23
THANK YOU
24

430634527-Final-Facial-Emotion-Detection-ppt-pdf.pdf

  • 1.
    (Roll no- 91/MCA/090021) Projectunder the guidance of DR. AMLAN CHAKRABARTI 1
  • 2.
    What is Emotion? Emotions are reflected in voice, hand and body gestures, and mainly through facial expressions. 2
  • 3.
    Facial Emotion  Thereare six types of facial emotions. Anger Fear Disgust Happy Sad Surprise 3
  • 4.
    Importance of emotionrecognition  Human beings express emotions in day to day interactions.  Understanding emotions and knowing how to react to people’s expressions greatly enriches the interaction. 4
  • 5.
    Review Of PreviousWork  Image pre-processing  Facial feature extraction -select some point on feature region -consider a matrix surrounding that point -calculate distances  Facial expression detection Reff.: [“Optical flow based analyses to detect emotion from human facial image data”- Axel Besinger , Tamara Sztynda, Sara Lal , Carmen Duthoit, Johnson Agbinya, Budi Jap ,David Eager, Gamini Dissanayake] 5
  • 6.
    Proposed Methodology  CaptureImage  Image Pre-processing -RGB to gray scale conversion -Scale Normalization  Crop Feature Regions  Edge Detection -Prewitt edge detection  Facial Emotion Classification 6
  • 7.
  • 8.
    Capture Image  Digitalcamera, web-cam  Resolution must be fixed  Distance between the web-cam and the face has to be fixed.  Position of face must be fixed  Must not wear glasses or have facial hair. 8
  • 9.
    Capture Image(Cont…) Normal FacialImage Emotional Facial Image 9
  • 10.
    RGB to grayscale conversion Color Facial Image Grayscale Facial Image  Reduce complexity. 10
  • 11.
    Scale Normalization  Selectonly the face portion within a rectangular frame  Remove unnecessary information  By defining region. 11
  • 12.
    Crop Feature Regions Region of Interest: two Eyes, two Eye brows, Lip  By defining the particular regions. 12
  • 13.
    Edge detection ROI ROIafter edge detection 13
  • 14.
    Algorithm: Input: A SampleImage Output: Detected Edges Step 1: Accept the input image Step 2: Apply mask h1 and h3 (prewitt mask) to the input image. Step 3: Masks manipulation of h1 and h3 separately on the input image Edge Detection (Cont…) 14
  • 15.
    Edge detection(Cont…) Step 4:Results combined to find the absolute magnitude of the gradient Where Step 5: The absolute magnitude is the output edges. 15
  • 16.
    Facial Emotion Classification Calculate displacement of the points.  Calculate averages of those values.  Calculate standard deviation (SD) of those averages.  Compare this SD with a threshold.  Get Emotion. 16
  • 17.
    Algorithm  Step 1:Take a still image of a normal expression pic1 (say) of a human face.  Step 2: Converts the color image to grayscale.  Step 3: Crop the five facial image region of interest (ROI) (eyes, eye brows and lip) from the image by defining region.  Step 4: Find edges of all image region.  Step 5: Take a still image of a emotional face (angry or happy) pic2 (say) of same person and repeat step 2, 3 and 4. 17
  • 18.
    Algorithm(Cont…)  Step 7:Comparing the deviation of edges of the specified region of pic1 with pic2 by finding the Euclidian distances of coordinate of each pixel.  Step 8: Put the Euclidian distances in a array k (say).  Step 9: Find the standard deviation (SD) from the elements of array.  Step 10: Comparing the SD with pre-define threshold and get the emotions. 18
  • 19.
    Experiment & Result 30 images were used to create the template.  60 were tested. Statistical Values of Three categories  19
  • 20.
    Experiment & Result Total 83.33% classify Facial Emotion. 63.33% were correct 20
  • 21.
    Experiment & Result 80% happy Emotion classification were correct.  50% Angry Emotion classification were correct.  Almost 100% null classification were correct. 21
  • 22.
    Applications  E-Learning system Robotics System  Human-Computer interaction System 22
  • 23.
    Conclusion and FutureWork  Implementation through this process is quite easy.  Have to improve over the capturing Process.  ROI extracted manually -That should be automatically.  Edge detection procedure should be less complex. 23
  • 24.