ALFA - Aparelho Lúdico para Fonoterapia e Alfabetização
Data
2017-12-11Autor
Fernandes, Maria Isabel Sousa
http://lattes.cnpq.br/7363280732490255
http://lattes.cnpq.br/1552297662479878
Meireles, Sara Silva
Metadados
Mostrar registro completoResumo
Studies in the area of Speech-Language Pathology have found ineffectiveness in the
methods used in the treatment of phonological problems in children. The main
obstacles are the shortage of tools and alternatives for the help of speech therapists,
educators and parents involved with children who present problems of this nature.
From this, the project Linguistic Apparatus for Speech Therapy and Literacy (ALFA),
comprised between the areas of Electronics, Informatics, Speech-Language
Pathology and Pedagogy; has as main objective to assist professionals in the field of
speech therapy in the treatment of people, especially children, who present such
changes, ie difficulty in articulation and pronunciation of speech sounds in the
Portuguese language, as well as those that are in the process of rehabilitation after
cochlear implantation. The project proposes a system consisting of an audio input
device whose abstractions make up a vector of descriptors that is passed to an
intelligent algorithm responsible for indicating the correctness of the phoneme. The
ALFA consists of four steps: research on the intelligence aspects of the system; data
capture for training, validation and machine learning tests; building a low-cost
microphone; and finally, the construction of the network itself. At the same time, the
ludic interface of the system is implemented. The results obtained with the
accomplishment of the first three stages were positive in relation to the acceptance of
the target public, the last tests with the KNN classification algorithm presented a
syllable accuracy rate between 0% and 60%, and both success in the construction of
the circuit of capture of audio of low cost, as in its tests in laboratory. From the results
achieved during the project development, we intend to perform new tests to define
the best machine learning technique, as well as to implement the ludic interface in
the system. In this way, it will be possible to test the system designed with at least
one patient in speech therapy and in a literacy class, proving its effectiveness.