Browse Prior Art Database

USE OF FACIAL EXPRESSION FLOW TO ANALYSE AND SCORE THE POSSIBILITY OF DEPRESSION

IP.com Disclosure Number: IPCOM000240572D
Publication Date: 2015-Feb-10
Document File: 9 page(s) / 516K

Publishing Venue

The IP.com Prior Art Database

Related People

S. Margret Anouncia: INVENTOR [+3]

Abstract

This paper proposes a context-based approach to recognize the flow of expressions while a person is talking thereby using those features to distinguish between healthy controls and depressed persons. Each frame has been analyzed based on the expression exhibited by the previous frames. This work has been accomplished by segmenting the video into framesets assuming that frames in a frameset exhibit the same expression. Segmenting the video into framesets has been accomplished in a dynamic manner based on the change of expressions. Hence this work focuses on three states of an expression (onset, peak and offset) too. A graph called Onset-Peak-Offset graph has been computed based on the state of an expression. This graph has been used effectively to distinguish the range of expressions depicted by a healthy control and a depressed person.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 14% of the total text.

Page 01 of 9

1

USE OF FACIAL EXPRESSION FLOW TO ANALYSE AND SCORE THE POSSIBILITY OF DEPRESSION.

Kuhelee Roy, G Subrahmanya VRK Rao, and S. Margret Anouncia

Abstract - This paper proposes a context-based approach to recognize the flow of expressions while a person is talking thereby using those features to distinguish between healthy controls and depressed persons. Each frame has been analyzed based on the expression exhibited by the previous frames. This work has been accomplished by segmenting the video into framesets assuming that frames in a frameset exhibit the same expression. Segmenting the video into framesets has been accomplished in a dynamic manner based on the change of expressions. Hence this work focuses on three states of an expression (onset, peak and offset) too. A graph called Onset-Peak-Offset graph has been computed based on the state of an expression. This graph has been used effectively to distinguish the range of expressions depicted by a healthy control and a depressed person.

---------- ----------

INTRODUCTION

 The inception of analysis of facial expression for emotion signaling can be attributed to the research on deciding upon the universality of a wide range of expressions across various cultures. Darwin, based on his studies on actions of facial muscles for exhibiting different emotion, claimed in one of his works that emotions are expressed by people across various cultures in almost similar ways. Analysis of facial expression finds its roots in various applications like speech illustration, conversation regulation, emblematic gestures, cognition, emotion signaling, expressive regulation etc.

 But the techniques mentioned do not take into account an effective description of the dynamics of a facial expression. This has been accomplished by segmenting an expression into onset, apex and offset states. Temporal segmentation of facial behavior is one more problem posed by real-world problems. This problem has been solved by a two-step process via, extraction of shape and appearance features followed by forming clusters of coherent facial gestures.

This work revolves around analysis of expressions for depressed patients and healthy controls. The major contribution of this paper is to analyze these motor characteristics on face to obtain a graph that can distinguish between the expressions flow of a healthy control to that of a depressed person. This has been envisaged by making use of computer vision algorithms for analyzing the flow dynamics of facial expressions. The expression analysis proposed in this paper takes into account the features extracted from each frame as well as the flow of expressions exhibited across a set of frames.


Page 02 of 9

7

OVERVIEW OF THE SYSTEM

  A novel method for analyzing expression flow has been proposed to distinguish between healthy controls and depressed persons. Hence the proposed algorithm encompasses a training-testing paradigm to analyze the flow of expressions for healthy controls....