Learning Conditional Acoustic Latent Representation with Gender and Age Attributes for Automatic Pain Level Recognition

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.

[1]  Jean Stains,et al.  Sex differences in regional brain response to aversive pelvic visceral stimuli. , 2006, American journal of physiology. Regulatory, integrative and comparative physiology.

[2]  Stefano Ermon,et al.  InfoVAE: Balancing Learning and Inference in Variational Autoencoders , 2019, AAAI.

[3]  Chi-Chun Lee,et al.  Toward Development and Evaluation of Pain Level-Rating Scale for Emergency Triage based on Vocal Characteristics and Facial Expressions , 2016, INTERSPEECH.

[4]  Panayiotis G. Georgiou,et al.  Behavioral Signal Processing: Deriving Human Behavioral Informatics From Speech and Language , 2013, Proceedings of the IEEE.

[5]  Li Fei-Fei,et al.  Tackling Over-pruning in Variational Autoencoders , 2017, ArXiv.

[6]  William Gutiérrez Lombana,et al.  Pain and gender differences: A clinical approach☆ , 2012 .

[7]  Michel F. Valstar,et al.  Fusing Deep Learned and Hand-Crafted Features of Appearance, Shape, and Dynamics for Automatic Pain Estimation , 2017, 2017 12th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2017).

[8]  Xing Zhang,et al.  Three dimensional binary edge feature representation for pain expression analysis , 2015, 2015 11th IEEE International Conference and Workshops on Automatic Face and Gesture Recognition (FG).

[9]  Jan Nielsen,et al.  Pain in the lumbar, thoracic or cervical regions: do age and gender matter? A population-based study of 34,902 Danish twins 20–71 years of age , 2009, BMC musculoskeletal disorders.

[10]  K. Craig,et al.  A theoretical framework for understanding self-report and observational measures of pain: a communications model. , 2002, Behaviour research and therapy.

[11]  Chi-Chun Lee,et al.  Embedding stacked bottleneck vocal features in a LSTM architecture for automatic pain level classification during emergency triage , 2017, 2017 Seventh International Conference on Affective Computing and Intelligent Interaction (ACII).

[12]  Simon Östlund,et al.  Women report higher pain intensity at a lower level of inflammation after knee surgery compared with men , 2017, Pain reports.

[13]  A. Stubhaug,et al.  Assessment of pain. , 2008, British journal of anaesthesia.

[14]  Lars Arendt-Nielsen,et al.  Age effects on pain thresholds, temporal summation and spatial summation of heat and pressure pain , 2005, Pain.

[15]  Donna Brown,et al.  Pain Assessment with Cognitively Impaired Older People in the Acute Hospital Setting , 2011, Reviews in pain.

[16]  Björn W. Schuller,et al.  The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for Voice Research and Affective Computing , 2016, IEEE Transactions on Affective Computing.

[17]  M. Robinson,et al.  The perception of pain in others: how gender, race, and age influence pain expectations. , 2011, The journal of pain : official journal of the American Pain Society.

[18]  Lars Arendt-Nielsen,et al.  Aalborg Universitet The effect of age and gender on pressure pain thresholds and suprathreshold stimuli , 2016 .

[19]  J. Zubieta,et al.  Gender and age influences on human brain mu-opioid receptor binding measured by PET. , 1999, The American journal of psychiatry.

[20]  R. C. L. Lima Júnior,et al.  A Clinical Experimental Model to Evaluate Analgesic Effect of Remote Ischemic Preconditioning in Acute Postoperative Pain , 2016, Pain research and treatment.

[21]  A. Unruh.  Gender variations in clinical pain experience , 1996, Pain.

[22]  James S Hodges,et al.  Evaluation of pretreatment analgesia and endodontic treatment for postoperative endodontic pain. , 2008, Journal of endodontics.

[23]  Kamal Nasrollahi,et al.  Deep Pain: Exploiting Long Short-Term Memory Networks for Facial Expression Classification , 2017, IEEE Transactions on Cybernetics.

[24]  Pieter Abbeel,et al.  Variational Lossy Autoencoder , 2016, ICLR.

[25]  Wolfgang Ellermeier,et al.  Gender differences in pain ratings and pupil reactions to painful pressure stimuli , 1995, Pain.

[26]  R. Helme,et al.  Age-related differences in pain perception and report. , 2001, Clinics in geriatric medicine.

[27]  Huachun Tan,et al.  Variational Deep Embedding: An Unsupervised and Generative Approach to Clustering , 2016, IJCAI.

[28]  M. Bullard,et al.  Revisions to the Canadian Emergency Department Triage and Acuity Scale (CTAS) adult guidelines. , 2008, CJEM.

[29]  M. Mccaffery,et al.  Improving the Quality of Care Through Pain Assessment and Management , 2008 .

[30]  R. Fillingim,et al.  Sex, gender, and pain: a review of recent clinical and experimental findings. , 2009, The journal of pain : official journal of the American Pain Society.