Title: Vision-Based Sign Language Recognition Systems for Low-Resource Languages: A Systematic Review with Focus on Tanzanian Sign Language
Authors: Isack Alex Masunzu, Salim Kassim Ally, Elipidius Itema Ladislaus, Eliabu Ntilagana, Devotha Kachere, Emmanuel Simwimba, Jeza Tunje
Volume: 10
Issue: 4
Pages: 112-119
Publication Date: 2026/04/28
Abstract:
Vision-based Sign Language Recognition Systems (SLRS) have advanced substantially over the past decade, yet this progress remains narrowly concentrated on a small cluster of well-resourced sign languages. Communities that rely on lesser-documented languages, including Tanzanian Sign Language (TSL), have been left without viable assistive technologies. This paper presents a systematic review of 28 peer-reviewed studies spanning sensor-based and vision-based SLRS approaches, with focused attention on the challenges specific to low-resource environments. We examine the full evolution of recognition methodologies from early handcrafted-feature pipelines to contemporary transformer-based architectures, analyze structural biases embedded in benchmark datasets, and critically evaluate the extent to which current evaluation practices account for real-world deployment constraints. Using TSL as a primary case study, we demonstrate how the dominant research paradigm characterized by large annotated corpora, GPU-intensive training, and centralized cloud inference is fundamentally misaligned with conditions prevalent across sub-Saharan Africa. From this analysis, we propose a four-component equitable SLRS development framework and derive a set of concrete, actionable recommendations for researchers, institutions, and policymakers toward genuinely inclusive and deployable sign language recognition solutions.