Meta-analysis of voice disorders databases and applied machine learning techniques

Voice disorders are pathological conditions that directly affect voice production. Computer based diagnosis may play a major role in the early detection and in tracking and even development of efficient pathological speech diagnosis, based on a computerized acoustic evaluation. The health of the Voice is assessed by several acoustic parameters. The exactness of these parameters is often linked to algorithms used to estimate them for speech noise identification. That is why main effort of the scientists is to study acoustic parameters and to apply classification methods that achieve a high precision in discrimination. The primary aim of this paper is for a meta-analysis on voice disorder databases i.e. SVD, MEEI and AVPD and machine learning techniques applied on it. This field of study was systematically reviewed in compliance with PRISMA guidelines. A search was performed with a set of formulated keywords on three databases i.e. Science Direct, PubMed, and IEEE Xplore. A proper screening and analysis of articles were performed after which several articles were also excluded. Forty-five studies that fulfills the eligibility criteria were included in this meta-analysis. After applying eligibility criteria on the peer reviewed and research article and studies that were published in authentic journals and conferences proceedings till June 2020 were chosen for further full-text screening. In general, only those articles that used voice recordings from SVD, MEEI and AVPD databases as a dataset is included in this meta-analysis. : We discussed the strengths and weaknesses of SVD, MEEI and AVPD. After detailed analysis of the studies including the techniques used and outcome measurements, it was also concluded that Support Vector Machine (SVM) is the most common used algorithm for the detection of voice disorders. Other than was also noticed that researchers focus on supervised techniques for the clinical diagnosis of voice disorder rather than using unsupervised techniques. It was also concluded that more work needs to be on voice pathology detection using AVPD database.