Design and Implementation of a Multi-Purpose Low-Cost Hall-Effect Sensor Glove for Sign Language Recognition

Dinanath Padhya, Jenish Pant, Krishna Acharya, Sajen Maharjan, Sudip Kumar Thakur · 2026

Read Design and Implementation of a Multi-Purpose Low-Cost Hall-Effect Sensor Glove for Sign Language Recognition by Dinanath Padhya, Jenish Pant, Krishna Acharya, Sajen Maharjan, Sudip Kumar Thakur (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-11539.txt.

Introduction

Dinanath Padhya and four coauthors build a low-cost sensor glove for sign language. Their 2026 preprint reports the design and implementation in full. Severe hearing loss affects over 430 million people worldwide. Access to sign language interpretation stays critically scarce.

The gap bites hardest in low-resource settings like Nepal. Assistive tools split into two flawed camps today. Commercial gloves often cost more than 3,000 dollars. Research gloves use flex sensors that degrade fast under stress. The team's glove uses Hall-effect sensors instead of flex ones. Hall sensors read magnetic fields as fingers move. The aim is a cheap, tough glove that lasts.

Catalog record

CREATORDinanath Padhya, Jenish Pant, Krishna Acharya, Sajen Maharjan, Sudip Kumar Thakur
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-11539

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Design and Implementation of a Multi-Purpose Low-Cost Hall-Effect Sensor Glove for Sign Language Recognition Thanks: Corresponding author: Dinanath Padhya Dinanath Padhya, Jenish Pant, Krishna Acharya, Sajen Maharjan, Sudip Kumar Thakur Abstract Despite the prevalence of severe hearing loss affecting over 430 million people globally, access to sign language interpretation remains critically scarce, particularly in low-resource settings like Nepal. Assistive technologies divide into two flawed categories: prohibitively expensive commercial gloves (often exceeding $3,000) or fragile research prototypes reliant on flex sensors that degrade rapidly under mechanical stress. This paper introduces a robust, cost-effective sign language recognition system tailored for the Nepali Sign Language (NSL) community. Departing from traditional resistive sensing, we implement a non-contact Hall-effect architecture that correlates magnetic field intensity with finger flexion, eliminating mechanical wear and signal drift. The system integrates 14 sensor nodes across the DIP, PIP, and MCP joints, augmented by an MPU6050 IMU for wrist orientation. An embedded Multi-Layer Perceptron, executed locally on an Arduino Mega, performs gesture classification, negating the need for cloud dependencies. With a Bill of Materials between $80 and $100, this solution is approximately 30 times more affordable than market alternatives. Validation trials across five subjects yielded 96% accuracy on a fundamental NSL vocabulary. As a proof-of-concept demonstration, this work validates the approach using 11 common NSL words, establishing a foundation for future vocabulary expansion. Stress testing confirmed that the Hall-effect configuration maintains signal fidelity over repeated cycles where traditional sensors fail. This study demonstrates that high-precision recognition is achievable through strategic engineering rather than premium components, offering a scalable pathway for deployment in Nepal’s deaf schools. Index Terms- Sign language recognition, Hall-effect sensors, wearable sensors, embedded neural networks, low-cost design I Introduction Hearing loss imposes a profound barrier to social integration, employment, and education for over 430 million people worldwide. In developing nations like Nepal, this exclusion is exacerbated by a chronic shortage of trained interpreters and educational resources. While the digital divide in urban centers is narrowing, rural access to sign language education remains virtually non-existent. Technological interventions have attempted to bridge this gap, yet current paradigms suffer from inherent structural flaws. Vision-based systems, while promising, demand significant computational resources and struggle with environmental variables such as poor lighting and occlusion. Conversely, wearable "smart gloves" offer a direct interface for gesture capture but are stifled by a dichotomy of cost and durability. Commercial solutions, such as the Bright Sign Glove, present a financial barrier so high (often >\$3,000 ) that they are irrelevant to the Global South. Meanwhile, the research landscape is saturated with prototypes using flex sensors [ 10 ] . These resistive strips are fundamentally ill-suited for long-term use; the act of bending causes material fatigue, leading to inevitable signal drift and device failure. This paper proposes a divergence from standard resistive sensing. We present the design and implementation of a durable, low-cost recognition glove specifically optimized for Nepali Sign Language (NSL). Our primary contribution is the architectural shift to non-contact Hall-effect sensors. By measuring magnetic flux rather than material deformation, we decouple the sensing mechanism from mechanical wear, ensuring long-term reliability at a fraction of the cost. The following sections detail the system’s electromechanical design, the on-board Artificial Neural Network (ANN) for static gesture classification, and validation results achieving 96% accuracy, proving that accessibility need not come at the expense of precision. II Related Work Engineers typically design recognition gloves by fusing flex sensors with Inertial Measurement Units (IMUs) [ 5 ] . This combination tracks finger bending and hand orientation. When paired with Artificial Neural Networks (ANNs), these systems achieve high accuracy for static gestures [ 5 , 4 , 7 ] . Results often exceed 98-99% for languages like American Sign Language (ASL) [ 5 ] . Flex sensors introduce a critical mechanical weakness. They operate as variable resistors. Bending degrades the resistive ink over time [ 6 , 3 ] . Researchers have tested alternatives to address this flaw. These include piezoelectric elements [ 6 ] , fiber optics [ 10 ] , knitted strain sensors [ 2 ] , and auxetic-interlaced yarn sensors [ 7 ] . This search for new materials confirms that standard resistive sensors lack the necessary durability for daily use. Cost presents a second barrier. Commercial hardware prices exclude most of the deaf community. Recent studies focus on accessibility. Prototyping with capacitive sensors and optimized electronics lowers expenses [ 9 , 8 ] . O’Connor et al. [ 8 ] and others [ 3 ] proved that sub-$100 designs are feasible. Our project targets this specific price benchmark. Linguistic bias also limits current technology. Most research focuses on American Sign Language (ASL) [ 5 , 10 ] . Regional languages receive little attention. Studies exist for Polish [ 6 ] , Korean [ 2 ] , and Indian sign languages [ 1 ] . Nepali Sign Language (NSL) remains largely undocumented in hardware studies. We address these three gaps simultaneously. We use durable Hall-effect sensors to solve the durability problem. These sensors are robust, provide high precision, and resist mechanical wear. Our design uses inexpensive sensors and an efficient ANN model [ 5 , 7 ] to meet the sub-$100 price point established by O’Connor et al. [ 8 ] . We validate the system specifically for the NSL community, providing a solution for a previously overlooked population. III Methodology III-A System Architecture Fig. 1: System architecture showing data flow from sensor acquisition to speech output The architecture shown in 1 operates as a four-stage pipeline starting with the Data Collection module, where Hall-effect sensors and an IMU capture finger flexion and wrist orientation. This data moves to the Control Unit, which executes Data Preprocessing and runs the embedded ANN Classification Model to identify the static gesture. The system then passes the classification index to the Transmission stage, which broadcasts the data via a Bluetooth Module. Finally, a Host Device such as a smartphone or PC receives the signal to execute the corresponding text-to-speech or control function. III-B 3D Structural Design The mechanical foundation of the system is a custom-designed 3D-printed exoskeleton, engineered to provide rigid and repeatable sensor placement while accommodating the natural kinematics of the human hand. The complete structural design is shown in Figure 2 . Fig. 2: 3D CAD model showing the glove’s structural design and sensor mounting points The design employs a segmented architecture with individual “sleeves” fitted to each phalanx of the fingers. These sleeves are dimensioned based on anthropometric data from globally reported finger measurements, ensuring compatibility across users with varying hand sizes. Each segment is designed to house a Hall-effect sensor and its corresponding magnet while maintaining minimal bulk and weight. The exoskeleton’s modular construction allows for independent fabrication and assembly of each finger module, facilitating repairs and customization. Strategic placement of mounting points ensures that the magnetic field gradient remains consistent across the full range of joint motion. The backplate of the structure provides a stable platform for mounting the processing electronics, IMU, and power system, while also serving as the primary attachment point to the user’s hand via adjustable straps. The current proof-of-concept ( 3 ) uses an Arduino Mega for rapid prototyping. Production versions could integrate a custom-designed

Read the complete text