Résumé:
The development of an Arabic-language speech-to-text system to control
smart devices has enormous potential to revolutionise the way we interact
with technology and enhance individuals' accessibility. This project
addresses the basic need for smooth communication between humans and
smart devices, especially for Arabic speakers. By integrating easy-to-use
components such as the microphone and the Raspberry Pi, users can
interact with the system effortlessly. They simply speak to the device using
a microphone to convert the speech to a text, and subsequently, the text is
used to control the smart devices or computers, as well as the text may be
used to navigate on the internet (especially in smart TVs). Our system is
based on the embedded computer (Raspberry Pi), and a deep learning
model powered by TensorFlow to transcript the input speech in real time.
By leveraging advanced speech recognition algorithms, the system
transcripts Arabic speech into accurate text. This innovative approach
overcomes the need for manual interactions or high cost devices (e.g. high
capacity smart TV). We particularly focus on low capacity computers
under Windows OS, as well as embedded devices with Android OS
(typically Smartphones).
