Computer Vision and Voice-Based AI Technologies for Assistive Systems

Abstract

Assistive technologies increasingly apply artificial intelligence to support access to physical environments, visual information, communication, and human-computer interaction. This review synthesizes eight selected studies covering indoor navigation, interactive object search, Indian Sign Language recognition, gesture-to-text interaction, real-time hand tracking, vision-language navigation, multilingual assistance, and wake-word detection. The studies are organized into two broad themes: computer vision-based assistance and voice-based AI. A narrative comparison examines the tasks addressed, technical approaches, input and output modalities, evaluation designs, reported findings, and limitations. The literature illustrates a movement from infrastructure-assisted guidance and constrained recognition toward open-vocabulary object search and vision-language interaction, while also showing the continuing importance of efficient perception and speech-activation components. The reviewed studies use heterogeneous evidence, including small user studies, image-classification experiments, device-level measurements, surveys, and prototype demonstrations; their results therefore should not be ranked through one common accuracy score. Recurring needs include broader real-world validation, generalization across users and environments, sustained contextual interaction, accessible user feedback, efficient deployment, and more consistent evaluation reporting. This paper is a synthesis of existing published work and does not propose a new system or report project-specific experiments. Keywords: Assistive Technology; Computer Vision; Voice-Based AI; Accessibility; Human-Computer Interaction; Multimodal AI