Multi-view representation learning via gcca for multimodal analysis of Parkinson's disease

• Published in: 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) pp. 2966 - 2970
• Main Authors: Vasquez-Correa, J. C., Orozco-Arroyave, J. R., Arora, R., Noth, E., Dehak, N., Christensen, H., Rudzicz, F., Bocklet, T., Cernak, M., Chinaei, H., Hannink, J., Nidadavolu, Phani Sankar, Yancheva, M., Vann, A., Vogler, N.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-03-2017
• Subjects: Correlation; Feature extraction; Frenchay dysarthria assessment; Gait processing; GCCA; Handwriting processing; Monitoring; Multi-view learning; Parkinson's disease; Speech; Speech processing; Training data; UPDRS
• ISSN: 2379-190X
• DOI: 10.1109/ICASSP.2017.7952700

Information from different bio-signals such as speech, handwriting, and gait have been used to monitor the state of Parkinson's disease (PD) patients, however, all the multimodal bio-signals may not always be available. We propose a method based on multi-view representation learning via generalized canonical correlation analysis (GCCA) for learning a representation of features extracted from handwriting and gait that can be used as a complement to speech-based features. Three different problems are addressed: classification of PD patients vs. healthy controls, prediction of the neurological state of PD patients according to the UPDRS score, and the prediction of a modified version of the Frenchay dysarthria assessment (m-FDA). According to the results, the proposed approach is suitable to improve the results in the addressed problems, specially in the prediction of the UPDRS, and m-FDA scores.