Abstract
BACKGROUND In Indonesia, people with communication disorders predominantly use Bahasa Isyarat Indonesia (BISINDO) as an alternative. Nevertheless, all current systems for BISINDO recognition generate fragmented, word-by-word output and lack natural language generation (NLG) capabilities, thereby fundamentally constraining their communicative utility. OBJECTIVE This study aims to develop and validate a prototype of an artificial intelligence (AI)-assisted wearable sign language-to-speech system—SmartGlove BISINDO—that embeds a large language model (LLM)- powered NLG into a BISINDO translation pipeline for the first time. METHODS Based on the Design Science Research (DSR) framework. This study follows the Design Science Research (DSR) framework, covering Phases 1--3: problem identification, solution objectives, prototype design, and development. Data were collected from 12 purposively selected informants, including sign language experts, AI technology expert linguists, and inclusive education practitioners in three cities in Indonesia (Malang, Surabaya, Makassar). Data were collected using a needs-analysis questionnaire and an expert validation sheet and analyzed through statistical triangulation of the agreement percentage, Content Validity Ratio (CVR), and Aiken's V. RESULTS The needs analysis led to a grand mean of 4.77 (SD = 0.51) across four clusters (Very High priority, 89.3% agreement). The proposed prototype focuses on the core limitation of all previous systems through a five-layer architecture that integrates dual-sensor wearable gloves (22 features/frame @ 100Hz), performs on-device TFLite LSTM inference for 150 BISINDO gesture classes, and interfaces with an LLM-based AI Language Wrapper for NLG. Expert validation demonstrated an overall mean Aiken's V of 0.94 across 14 items, with 11 (78.6%) items categorized as Highly Valid, and a grand mean CVR of 0.77, exceeding the Lawshe minimum threshold. CONCLUSIONS SmartGlove BISINDO is a novel sign language-to-speech system that introduces the BISINDO research landscape to LLM-based NLG connect-and-translate approaches and demonstrates strong content validity. This study presents a replicable model for interdisciplinary DSR-based prototype development that combines AAC theory, Systemic Functional Linguistics, and design science methodology, and can be applied to other underrepresented sign language systems found in Southeast Asia and elsewhere.

This publication has 32 references indexed in Scilit: