University of Southern Queensland
Research Field: Computer Science, Chronic Disease
The key aims of Dr Rana’s research are two-fold: 1) determine “mood” automatically from spontaneous phone conversations and 2) determine the onset of relapse from mood patterns. To determine the mood of person A, the discrete emotions in a caller’s/callee’s (person B and others) voice will be used as an eliciting probe and the emotional response of A will be used to evaluate A’s mood. When in a positive mood, A should respond appropriately to B’s emotions, i.e., express sympathy/sadness to sorrow, express excitement/happiness to humour, etc. The underlying research challenges are as follows: 1) separate A’s speech from that of B’s , 2) extract emotions (e.g., happiness, sadness etc.) of A and B’s speech, 3) map emotions of A with that of B, 4) quantify the “match/miss match” to label A is in a positive/negative mood, and 5) maintain privacy by not listening/recording the calls.
To predict relapse, Deep Learning algorithms are being developed upon training with information on mood and clinical diagnosis of relapse collected from patients. The end product will be an application on A’s phone that will be able to flag an onset of relapse to A and his/her clinician – just based on day-to-day phone calls.