Rajeev Galaiya Undergraduate Dissertation 2017/18

Matching Faces to Voices

Supervised by R.K.Moore

Abstract

Artificial agents have been a figment of the imagination for a long time, with everybody having different opinions on how they would look and sound like. In the last few years, the growth of humanoid robots has meant that this is now a reality. The purpose of this project is to investigate the relationship between faces and voices. The paper looks at some of the key factors that humans use to match a face to a voice and vice versa. An online survey was conducted where participants were asked to match faces and voices and give reasons as to why they chose their combination. Some of the voices and faces were manipulated to look and sound more robotic, and participants were asked to rate these using different criteria such as pleasantness, naturalness and authoritativeness. Results from the survey revealed that most people use the size of the face to match different voices. Smaller faces are more associated with higher pitch voices while longer and broader faces are more associated with lower pitch voices. Another factor was skin color as many participants mentioned that they associated darker skin complexion with deeper voices