Rehabilitation support system for speech disorder by acoustic analysis

2022 Fiscal Year Final Research Report

• Project/Area Number: 19K20751
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 90150:Medical assistive technology-related
• Research Institution: University of Hyogo (2022); Himeji Dokkyo University (2019-2021)
• Principal Investigator: Yagi Naomi 兵庫県立大学, 先端医療工学研究所, 准教授 (40731708)
• Project Period (FY): 2019-04-01 – 2023-03-31
• Keywords: 発話障害 / リハビリテーション / 人工知能 / 簡易診断システム

Outline of Final Research Achievements

In this study, the algorithm was examined mainly for acoustic analysis for the purpose of creating a rehabilitation support tool for dysarthria. To compare between the groups of healthy subjects and dysarthria patients, we focused on the fluctuation of the period and amplitude of the speech signal during vocalization. Using fluctuation measurement of speech data, we predicted a model for detecting utterance anomalies based on features such as fluctuation index, fluctuation of YURAGI, and dynamic time warping. A weighted inverse propensity score matching method using the reciprocal of the propensity score was applied to remove inter-group variability. As the results, the area under the classification performance curve for healthy subjects and patients was 0.78. Acoustic analysis techniques have been shown to be useful in diagnosing and treating dysarthria.

Free Research Field

医工学

Academic Significance and Societal Importance of the Research Achievements

医療現場では、エビデンスに基づく臨床判断が求められる。現在、発話障害検査は、話しことばを聴収し、その聴覚印象によって行うため、評価は検査者の主観に必然的に依存する。そのため、評価結果に個人差が生じることもあり、経験に基づいた高い専門的能力が求められると共に精度や信頼性の点で検討の余地が残る。本システムでは、場所、時間、体位・動作等の制限や拘束が少なく、さらに専門的な技術が不要な、安価で簡便な検査を実現可能となった。機能不全等の見える化を図ることでリハビリテーションを効率良く実施できたと考える。