Electronic Theses and Dissertations

Identifier

987

Date

2013

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Engineering

Concentration

Computer Engineering

Committee Chair

MOHAMMED YEASIN

Committee Member

Russel Deaton

Committee Member

Bashir Morshed

Committee Member

Arthur Graesser

Abstract

This dissertation presents the design and implementation of EmoAssist: "Emotion-Enabled Assistive Tool to Enhance Dyadic Conversation for the Blind". The key functionalities of the system are to recognize behavioral expressions and to predict 3-D affective dimensions from visual cues and to provide audio feedback to the visually impaired in a natural environment. Prior to describing the EmoAssist, this dissertation identifies and advances research challenges in the analysis of the facial features and their temporal dynamics with Epistemic Mental States in dyadic conversation. A number of statistical analyses and simulations were performed to get the answer of important research questions about the complex interplay between facial features and mental states. It was found that the non-linear relations are mostly prevalent rather than the linear ones. Further, the portable prototype of assistive technology that can aid blind individual to understand his/her interlocutor's mental states has been designed based on the analysis. A number of challenges related to the system, communication protocols, error-free tracking of face and robust modeling of behavioral expressions /affective dimensions were addressed to make the EmoAssist effective in a real world scenario. In addition, orientation-sensor information from the phone was used to correct image alignment to improve the robustness in real life deployment. It was observed that the EmoAssist can predict affective dimensions with acceptable accuracy (Maximum Correlation-Coefficient for valence: 0.76, arousal: 0.78, and dominance: 0.76) in natural conversation. The overall minimum and maximum response-times are (64.61 milliseconds) and (128.22 milliseconds), respectively. The integration of sensor information for correcting the orientation has helped in significant improvement (16% in average) of accuracy in recognizing behavioral expressions. A user study with ten blind people shows that the EmoAssist is highly acceptable to them (Average acceptability rating using Likert: 6.0 where 1 and 7 are the lowest and highest possible ratings, respectively) in social interaction.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Share

COinS