Implementation and Design of Luganda voice-controlled Elevator System.

Abstract

In recent years, voice-activated systems have become a cornerstone of modern human-computer interaction, yet many such systems lack support for indigenous languages like Luganda. This project focuses on the implementation and design of a voice-controlled elevator system that understands and responds to spoken commands in Luganda, one of Uganda’s major local 9 languages. Leveraging machine learning and natural language processing (NLP) techniques, the system performs intent recognition to interpret user commands such as requesting specific floors, opening or closing the elevator door. The proposed solution integrates a Luganda speech-to-intent pipeline, combining automatic speech recognition (ASR) with a supervised intent classification model. Given the low-resource nature of Luganda, a custom dataset was developed through audio data collection and manual annotation. The model was trained to recognize key elevator-related intents and deployed on an embedded system equipped with a microphone interface. The result is a functional prototype capable of facilitating hands-free, voice-based control of an elevator system in a local language.