I am currently working as a Research Scientist at Pearson Knowledge Technologies. For my current website click here.
Mail : 3740 McClintock Ave. EEB 400
Los Angeles, CA 90089-2560
Office : RTH 318
Phone: (213) 821-4282
e-mail : metallin at usc dot edu
I am a PhD
candidate in the Electrical Engineering (EE) department
of the University of Southern California (USC). My
advisor is prof. Shrikanth Narayanan. I am a member
of the Speech Analysis and Interpretation Laboratory (SAIL) since September 2007. I am graduating in Spring 2013.
My research interests include multimodal and speech signal processing, machine learning, statistical modeling and
affective computing. My thesis topic focuses on developing accurate computational models
for emotion recognition and understanding. Such models take into account the context in
which an emotion is expressed, and capture the evolution of an emotional expression
throughout interactions, highlighting regions of interest. Multimodality is a
fundamental aspect of emotional expression and my work studies a variety of
emotion-modulated modalities such as facial expressions, vocal characteristics, as well as
body language movements and gestures conveying interaction information between
participants of an affective conversation. Finally my thesis explores applications of such computational approaches in healthcare, specifically for analysis of affective facial expressions of children with autism spectrum disorders
This type of research has given me the unique opportunity to get involved with
exciting multi-disciplinary collaborations with experts from domains quite different from
engineering, such as theater and psychology.
In parallel, I have also worked in the speech and dialog systems area, as part of my microsoft research internship, where we explored the use of machine learning approaches for discriminative belief tracking for spoken dialog systems.
This website provides information about my research progress and
publications. Please don't hesitate to contact me for comments or questions.