Otimização do sistema de processamento de áudio da plataforma eFono

Abstract

The phonological disorder is defined as a speech sound disorder characterized by the inadequate production of sounds and inadequate use of the phonological rules of the language. In this context, the eFono application appears, designed to assist speech-language therapists in the speech evaluation process. To this end, audios of patients pronouncing keywords are recorded by the professional, with the help of the application. Using eFono, problems were noticed in sound processing, such as slowness and unwanted sounds. This work proposes to improve the eFono audio acquisition process, optimizing the processing and classification of the pronunciation of words spoken by children. An approach that allows audio clipping in the eFono Android application was developed. In addition, a research was carried out regarding the sampling rate of the frequency of recorded audios, carrying out tests with different samples seeking to find the sampling rate without loss of quality, with the smallest size. The results achieved through the development of the clipping activity provide a reduction in the sizes of sound files, thus reducing the time of sending the audios to the server and their storage. Furthermore, the development of the frequency quality test flow facilitates the execution of future tests, which will be able to benefit from a larger sample base.