Investigation of Voice Pathology Detection and Classification on Different Frequency Regions Using Correlation Functions

Abstract

The best achieved accuracies in both detection and classification were varied depending on the band, the correlation function, and the database. The most contributive bands in both detection and classification were between 1000 and 8000 Hz. In detection, the highest acquired accuracies when using cross-correlation were 99.809%, 90.979%, and 91.168% in the Massachusetts Eye and Ear Infirmary, Saarbruecken Voice Database, and Arabic Voice Pathology Database databases, respectively. However, in classification, the highest acquired accuracies when using cross-correlation were 99.255%, 98.941%, and 95.188% in the three databases, respectively.

Towards robust voice pathology detection

“…Moreover, voice can be nowadays easily recorded using a variety of smart devices, and processed remotely using cloud technologies. From these reasons, works such as [17, [18,39], SVD -Saarbruecken Voice Database [62,44,2], AVPD -Arabic Voice Pathology Database [41,44], KM -K-means [23], RF -Random Forests [11], GMM -Gaussian Mixture Models [50], SVM -Support Vector Machines [24], NB -Naive Bayes [45], ELM -Extreme Learning Machine [30], and ANN -Artificial Neural Networks [53]. 26,44,2] focused on using signal processing techniques (to quantify vocal-manifestations of the pathology under focus) and machine learning algorithms (to automate the process of voice pathology detection) to build a system capable of accurate discrimination of healthy and pathological voices.…”

“…[36,17,26], etc. From the voice pathologies point of view, most researchers restricted the dataset to a limited set of pathologies [7,43,14,25,51,44,5,3,4,2].…”

“…To propose results comparable with the previously published works, we analyze voice recordings of sustained phonation of the vowel /a/ as well. However, unlike the previous works, we analyze a larger dataset composed of 4 different databases, namely: MEEI [18,39], SVD [62,44,44,2], AVPD [41,44] (these databases are commonly used by the community), and Príncipe de Asturias Database (PDA) [20,8,39]. Furthermore, to propose models capable of robust voice pathology detection, we do not restrict the dataset to only a subset of common vocal pathologies.…”

“…By using 4 different databases, we aim to increase the size of our dataset to enable exploring possibilities of using supervised deep learning techniques that de- [41,44], b) MEEI [18,39], PDA [20,8,39], and SVD [62,44,2].…”

Multilingual evaluation of voice disability index using pitch rate

“…Similar studies conducting comparative analyses of healthy individuals and patients have been performed using the same samples [22,23]. However, those studies utilized voice samples of the long-vowel sound "ah."…”