Edge-native static American Sign Language recognition using a lightweight geometric multilayer perceptron

TransSLR: A Lightweight Transformer for Sign Language Recognition

Automated Sign Language Recognition for under-represented languages remains a largely unsolved problem. Central African Sign Language (CASL) exemplifies this gap: the only available bench-mark, CASL-W60, has a best reported accuracy of 69.93%, and we show that the common heuristic of fine-tuning high-resource models fails to close it. This failure stems from two compounding factors: the limited scale of available CASL data and the significant lexical and visual domain gap between CASL and large-scale corpora such as WLASL, which renders pre-trained representations largely uninformative. To address this, we propose TransSLR, a lightweight Temporal Transformer Encoder trained from scratch on 64-frame normalized pose sequences, with average pooling and a classification head. By operating on geometric keypoint representations rather than raw RGB, TransSLR achieves signer-independent generalization without relying on visual appearance. On the CASL-W60 benchmark, TransSLR establishes a new state-of-the-art accuracy of 80.39%, surpassing the prior best by +10.46%. Beyond accuracy, our encoder-only design significantly reduces computational overhead, making deployment feasible in resource-constrained environments. We conduct extensive experiments on the CASL-W60 benchmark, comparing against RGB-based and multimodal baselines, and demonstrate that TransSLR achieves state-of-the-art performance.

A Framework for Lightweight, Edge-Based Recognition of Dynamic American Sign Language Using Temporal Learning Models

Automated recognition of dynamic American Sign Language (ASL) gestures remains a significant challenge for real-time deployment on resource-constrained edge devices. Although recent advances in deep learning have achieved high accuracy in sign language recognition systems, such approaches typically rely on GPU acceleration and substantial computational resources, limiting their feasibility for accessible, real-world applications.This study proposes a theoretical and methodological framework for evaluating lightweight machine learning models for dynamic ASL recognition under CPU-dependent constraints. Grounded in Human-Computer Interaction Theory, Multimodal Communication Theory, and Computational Learning Theory, the framework formalizes the relationship between temporal representation, model complexity, and computational feasibility in edge-based environments.The proposed framework outlines a comparative evaluation strategy using pose-based time-series data extracted from glossed-annotated ASL videos, examining both sequence-preserving models (e.g., Canonical Interval Forest and InceptionTime) and aggregated-feature classifiers (e.g., Random Forest and Logistic Regression). Rather than reporting empirical findings, this paper establishes the conceptual foundations, modeling assumptions, and evaluation criteria necessary to determine whether lightweight classifiers can approximate the performance of deep learning approaches while remaining suitable for edge deployment.By explicitly linking theoretical principles to methodological design choices, this work provides a foundation for future empirical studies and contributes a structured approach for developing accessible, efficient, and scalable sign language recognition systems.

A landmark-based temporal pipeline for real-time adjective recognition in Indian sign language

Optimizing MNIST Digit Recognition for Edge Devices: A Multi-dimensional Evaluation Approach

The authors introduce the Edge Suitability Score (ESS), a composite metric that combines normalized accuracy, model size, and inference speed into a single value, weighted at 0.40, 0.35, and 0.25 to reflect their relative importance for microcontroller deployment.

A Low-Cost Hybrid Reservoir Computing Model for Isolated Sign Language Video Recognition

Sign language recognition (SLR) enhances communication between hearing and hearing-impaired individuals. Although deep learning (DL) has achieved promising performance in SLR, its high computational cost limits deployment on edge devices. To address this challenge, we propose a lightweight reservoir computing (RC)-based approach for SLR. In the proposed method, MediaPipe extracts body and hand keypoints to capture the spatial and temporal dynamics of gestures. These keypoints are then processed by a hybrid reservoir computing (HRC) architecture that combines deep reservoir computing (DRC) and bidirectional reservoir computing (BRC), transforming the input into a high-dimensional dynamic representation. A ridge regression model maps the final HRC state to class labels. This HRC-based SLR method achieved Top-1, Top-5, and Top-10 accuracies of 61.12%, 86.05%, and 92.56%, respectively, on the Word-Level American Sign Language 100 (WLASL100) video dataset, demonstrating competitive performance compared to deep learning-based approaches. Additionally, due to the lightweight nature of RC, the training time was drastically reduced to only a few seconds compared with DL-based methods such as Bi-GRU.This method offers low computational cost, showing its potential for deployment on edge devices.

Lightweight AI-Driven Real-Time Swahili Sign Language Recognition: A Way to Empower Hearing-Impaired Individuals

Hearing impairment is a growing global health challenge that disproportionately affects people in resource-constrained countries such as Tanzania, where access to assistive technologies remains limited. Conventional solutions, including hearing aids, cochlear implants, and interpreters, are often costly and inaccessible. This study proposes an AI-driven real-time Swahili Sign Language Recognition system designed to bridge communication gaps between the hearing-impaired community and the general population. The system integrates Convolutional Neural Networks (CNNs), Scale-Invariant Feature Transform (SIFT), and Long Short-Term Memory (LSTM) with computer vision techniques to recognize and interpret sign gestures. SIFT extracts key features from images captured via webcam, while CNN performs classification and LSTM models temporal dependencies for improved accuracy. Experimental results reveals high recognition accuracy of 98.5% and a low error rate of 0.1345%, outperforming several existing models. The system offers an effective, affordable solution for enhancing communication among Swahili-speaking communities.