Sistema de identificação digital e refatoração aplicadas em um aplicativo para avaliações fonológicas

Abstract

Speech disorders are problems characterized by phonological processes that affect kids until their pre-adolescence. In this cases, the speech terapists perform evaluations and, if necessary, future therapies aiming to correct them. This scenario motivated the development of a mobile app, called eFono, to perform childish phonological screening; the app, however, still has some remaining problems like the impossibility of identification and grouping of evaluations performed by the same patient, the depreciated techonologies of the app and the restricted use of a selected group of users. To solve these problems, this work refactored the app source code and investigated a voice digital identification system to integrate with the app, choice motivated by the backwards compatibility with the voices already collected in the previous work and a combination of factors such as low computational cost, ease of use and high assertivity when compared to other known biometric identification systems. In the methodology, this work compares the results obtained from two libraries, Recognito and Alizé, which both perform the speaker identification, using a dataset that contains 1045 phonological evaluations. With the tests results, it was concluded that Recognito tool was not efficient enough to integrate with the app, whereas Alizé got an assertivity rate of 91.20%. To future projects, it is recommended the exportation of Alizé to a programming language that can be maintained in a server to realize the effective integration with the app.