Estimación del parámetro soplosidad dentro del método de análisis perceptual de la voz cape-v

Abstract

The voice is fundamental element of human communication, through it people transmit their feelings, emotions and ideas. In this context, different branches of medicine and engineering have opened upto the study of this phenomenon to explain in a way or other as it originates and the possible alterations, to detect these disorders of the voice the functional evaluation is used, To carry out this evaluation, the speech pathologist must apply different tests and procedures in order to assess each of the elements that intervene in the production of the spoken and sung voice. These elements are evaluated largely through auditory perceptual evaluation, palpation and visual observation, which makes this examination subjective, and with this that the result depends on the experience of the professional I evaluated. There are different scales that seek to standardize the perceptual assessment of the voice, such as GRABS and CAPE-V, but these still have a high level of subjectivity. Several studies have been developed to reduce professional subjectivity in this test, but most are developed around the GRABS scale, which has certain disadvantages. With the purpose of reducing the exam time and the subjectivity of the test, the development of an app that allows predicting the breathiness parameter within the CAPE-V perceptual evaluation method is proposed, through digital processing of acoustic signals and machine learning. To arrive at the final development of the App, a synthetic voice bank was created, which was used to create a dataset, which was implemented in attribute selection and regression models, in the Matlab and Weka softwares, obtaining with this the most optimal regression model for the prediction of soplocity.