Research Article

Ensemble Convolution Neural Network for Robust Video Emotion Recognition Using Deep Semantics

Algorithm 2

Template-based face detection in keyframes.
Input: the keyframe() and template image (t)
Output: The keyframe with face
Read the keyframes
Read the template image t
Apply template matching to detect face image
 Slide the t over
 Compare t and and find the correlation value (cv) using equation (1)
 Normalise the correlation value using (2)
 Compute the correlation threshold (T)
  Add mean with an arbitrary number of standard deviation
 Compare cv and T
 If the cv > T the
  The segment is marked as a face
 Else
  The segment is marked as nonface
 End If