An Intelligent Decision-Support Framework Using Mobile and Virtual Reality Technologies for Optimising Intangible Cultural Heritage Management

Abstract

Effective administration of intangible cultural heritage (ICH) increasingly faces challenges arising from intricate prioritization demands, constrained resources, and sustainability obligations. Conventional approaches to ICH management are predominantly manual and rely heavily on expert judgement, limiting their scalability and their ability to integrate data across multiple modalities. This study introduces an intelligent decision-support framework that incorporates interactive mobile interfaces, virtual reality (VR), and a novel Adaptive Convolutional Synthesized Intelligent Memory Network (ACSIMN) to facilitate precise ICH prioritization and optimise resource distribution through adaptive learning of multimodal features and cross-modal reasoning. Data are gathered from diverse sources, including images and videos of cultural artefacts, audio recordings of traditional music and oral histories, and textual materials detailing historical and cultural contexts. Modality-specific pre-processing procedures are applied, such as resizing visual inputs for images and videos, generating spectrograms from audio signals, and tokenising and cleansing textual information. Feature extraction is conducted using Scale-Invariant Feature Transform (SIFT) for visual inputs, Mel-Frequency Cepstral Coefficients (MFCCs) for audio data, and Bidirectional Encoder Representations from Transformers (BERT) for text. These features are then integrated into a unified latent space, enabling coherent multimodal correlation and informed joint decision-making. The principal innovation resides in the ACSIMN model, which harmonises adaptive convolutional neural network (ACNN) learning for visual interpretation with intelligent long short-term memory (ILSTM) for temporal and semantic sequence modelling, complemented by VR-enabled visualisation to improve situational comprehension. Implemented in Python, the proposed framework attains an accuracy of 98.4%, supporting both accurate recognition and real-time prioritization of cultural assets, while demonstrating potential for sustainable, scalable, and data-driven preservation of ICH.