Sistema in-door de monitoramento e controle para acessibilidade de pessoas com deficiência pelo reconhecimento de fala utilizando Raspberry Pi

Abstract

The evolution of technologies used in residential automation has raised several discussions about their use in favor of accessibility, since people with disabilities or reduced mobility would greatly benefit from the implementation of such technologies. Among the emerging mechanisms, the speech command interface stands out, as it allows some motor actions to be performed using the voice. In this context, this project aims to address the development of an in-door system with a command interface via speech, which aims to control domestic devices, such as television, air conditioning and lamps, to help people with reduced mobility. The Raspberry Pi development platform is used, associated with the infrared signal transmission circuit and the lamp switching relay. These modules are managed by a script developed in the Python programming language, making use of the Google Speech API. To send commands via infrared signals, the LIRC library is used. The system was evaluated for the efficiency of speech recognition when commands are pronounced at 10, 60 and 100 cm away from the microphone, presenting metrics of 100%, 87% and 73%, respectively, with an overall average of 87% of correct conversions. Aspects that influence the drop in efficiency in command recognition are discussed, and solutions for this problem are presented. Despite this factor, the proposal to control the devices was fulfilled, with the possibility of expanding it by including new means of activating different devices in a residence.