Title: Analysis of voice for differentiating Parkinson's disease affected persons from healthy persons
Authors: Vikas; R.K. Sharma
Addresses: M.D. University Campus, H.No-28/9J houses, Rohtak, Haryana-124001, India ' School of VLSI Design and Embedded Systems, NIT Kurukshetra, Haryana-136119, India
Abstract: Parkinson's disease (PD) is said to occur due to ups and downs in brain chemical named as 'dopamine'. Though there is no permanent cure of it yet some therapies provide relief to the PD affected persons. However, these therapies are too costly and beyond the reach of large number of patients. Approximately 90% of PD affected persons suffer from speech disturbances. In this paper, an attempt has been made for early detection of PD using various parameters of voice like pitch, Mel-frequency cepstral coefficients (MFCC), jitter, shimmer, intensity, glottal pulse and dynamic time warping technique (DTW). The parameters of the voice of PD affected persons were compared with healthy persons. Results revealed that the pitch, jitter, shimmer and 1st MFCC coefficient has large values in case of PD affected persons than healthy persons whereas the healthy persons were found to have larger values of intensity than PD affected persons. DTW-based feature matching was found more in case of PD affected persons than healthy persons. Implication of the finding for the PD affected person and health professionals are discussed.
Keywords: voice pitch; intensity; jitter; Mel-frequency cepstral coefficients; MFCC; glottal pulse; dynamic time warping; shimmer; voice analysis; Parkinson's disease; speech disturbances; feature matching.
International Journal of Telemedicine and Clinical Practices, 2015 Vol.1 No.2, pp.174 - 185
Received: 11 Jun 2014
Accepted: 23 Jul 2014
Published online: 09 Jun 2015 *