Pain is an unpleasant internal sensation caused by bodily damages or physical illnesses with varied expressions conditioned on personal attributes. In this work, we propose an age-gender embedded latent acoustic representation learned using conditional maximum mean discrepancy variational autoencoder (MMD-CVAE). The learned MMD-CVAE embeds personal attributes information directly in the latent space. Our method achieves a 70.7% in extreme set classification (severe versus mild) and 47.7% in three-class recognition (severe, moderate, and mild) by using these MMD-CVAE encoded features on a large-scale real patients pain database. Our method improves a relative of 11.34% and 17.51% compared to using acoustic representation without age-gender conditioning in the extreme set and the three-class recognition respectively. Further analyses reveal under severe pain, females have higher maximum of jitter and lower harmonic energy ratio between F0, H1 and H2 compared to males, and the minimum value of jitter and shimmer are higher in the elderly compared to the non-elder group.

Pain is an internal construct with vocal manifestation that varies as a function of personal and clinical attributes. Understanding the vocal indicators of pain-levels is important in providing an objective analytic in clinical assessment and intervention. In this work, we focus on investigating the variability of voice quality as a function of multiple clinical parameters at different pain-levels, specifically for emergency room patients during triage. Their pain-induced pathological voice quality characteristics are naturally affected by an individual attributes such as age, gender and pain-sites. We conduct a detailed multivariate statistical analysis on a 181 unique patient's vocal quality using recordings of real triage sessions. Our analysis show several important insights, 1) voice quality only varies statistically with pain-levels when interacting effect from other clinical parameters is considered, 2) senior group shows a higher value of voicing probability and shimmer when experiencing severe pain, 3) patients with abdomen pain have a lower jitter and shimmer during severe pain that is different from patients experiencing musculoskeletal pathology, and 4) there could be a relationship between the variation in the voice quality and the neural pathway of pain as evident by interacting with the pain-site factor.