基于AI方言数字人的适老化用药服务系统研发与应用研究

Development and Application of AI Dialect Digital Human-Based Elderly Medication Service System

摘要: 我国已进入深度老龄化社会，老年群体用药服务需求持续攀升，而方言区老年人语言使用习惯与医疗信息系统普通话通用标准之间的沟通壁垒，成为制约老年用药服务可及性的重要因素；与此同时，以浙江淳安话为代表的诸多地方小众方言，正面临代际传承断裂、生存空间萎缩的濒危困境。针对上述两大社会与文化问题，本研究选取淳安话为研究载体，设计并研发基于AI方言数字人的适老化用药服务系统。技术实践中，首先搭建面向用药场景的淳安话专项小型语料库，涵盖药品名称、用法用量、不良反应等医疗核心术语；其次依托WeNet语音识别与VITS语音合成框架，采用迁移学习方法对预训练模型进行微调，并结合淳安话音系特征构建规则增强模块；进而开发多模态AI数字人交互终端，实现药品说明书拍照识别、方言语音问答、用药定时提醒等核心功能；最后通过社区试点应用，初步验证系统的实际可用性与服务有效性。本研究同步构建淳安方言“语音–文本–文化语境”三维标注数据库，为小众濒危方言的数字化保护与场景化应用提供了参考性技术路径与实践经验。

Abstract: China has entered a deeply aging society, and the elderly group has an increasing demand for medication services. However, the communication barrier between the language usage habits of the elderly in dialect areas and the Mandarin general standard of medical information systems has become an important factor restricting the accessibility of medication services for the elderly. At the same time, many local minor dialects represented by Chun’an dialect in Zhejiang Province are facing an endangered dilemma of intergenerational inheritance rupture and shrinking living space. Aiming at the above two social and cultural problems, this study selects Chun’an dialect as the research carrier to design and develop an elderly-friendly medication service system based on AI dialect digital humans. In technical practice, firstly, a special small corpus of Chun’an dialect for medication scenarios is built, covering core medical terms such as drug names, usage and dosage, and adverse reactions. Secondly, based on WeNet speech recognition and VITS speech synthesis frameworks, the pre-trained model is fine-tuned by transfer learning, and a rule enhancement module is constructed combined with the phonological characteristics of Chun’an dialect. Furthermore, a multimodal AI digital human interaction terminal is developed to realize core functions such as medicine instruction photo recognition, dialect voice Q&A, and timed medication reminders. Finally, the practical usability and service effectiveness of the system are initially verified through community pilot applications. This study also constructs a 3D annotated database of Chun’an dialect integrating “speech, text and cultural context”, providing a reference technical path and practical experience for the digital protection and scenario-based application of minor endangered dialects.