Development of Arabic speech-to-text system for smart device control

Abstract

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).