| 講演抄録/キーワード |
| 講演名 |
2017-12-13 02:55
音声指標によるパーキンソン病患者とうつ病患者と健常者の鑑別 ○篠原修二(東大)・大宮康宏(PST)・中村光晃・樋口政和(東大)・萩原直樹・高野 毅(PST)・光吉俊二・徳野慎一(東大) |
| 抄録 |
(和) |
これまで我々は,健常者とうつ病患者,健常者とパーキンソン病患者を識別し得る音声指標の開発を進めてきた.これら二つの音声指標Major Depression Index (MDI) とParkinson Disease Index (PDI)は,音声波形のゼロ交差率とハースト指数の関係性に基づいて導出される.本研究の目的は,MDIとPDIを用いて,うつ病患者とパーキンソン病患者を識別可能か否か検討することである.分析には,健常者,パーキンソン患者,うつ病患者という三種類の被験者から取得した17種類の定型文読み上げ音声を用いた.分析の結果,MDIはすべてのデータから,うつ病患者音声のみを識別し得る傾向を示した.同様にPDIはすべての音声データから,パーキンソン病患者音声のみを識別し得る傾向を示した.以上の結果から,MDIとPDIはともにゼロ交差率とハースト指数から導出される指標ではあるが,異なる音声特徴を捉えた独立の指標と考えられ,両指標を用いて,健常者,うつ病患者,パーキンソン病患者の三者を鑑別できる可能性が示唆された. |
| (英) |
We have developed vocal indices that can discriminate between healthy individuals and patients with depression and between healthy individuals and patients with Parkinson's disease. These two vocal indices, that is Major Depression Index (MDI) and Parkinson Disease Index (PDI) are derived based on the relationship between the zero crossing rate and the Hurst exponent. The purpose of this study is to investigate whether MDI and PDI can be used to distinguish between depressed patients and Parkinson's patients. For the analysis, 17 kinds of fixed phrase reading sounds obtained from three kinds of subjects, healthy individuals, Parkinson patients, and depressed patients were used. As a result of the analysis, MDI showed a tendency to distinguish only depressed patient voice from all data. Likewise, PDI showed a tendency to distinguish only speech of Parkinson's disease patients from all voice data. Both MDI and PDI are indicators derived from the zero-crossing rate and Hurst exponent, but they are thought to be independent indices capturing different vocal features, and by using both indices, healthy subjects, depressed patients, and Parkinson's disease patients could be distinguished. |
| キーワード |
(和) |
ハースト指数 / ゼロ交差率 / パーキンソン病 / うつ病 / 音声指標 / 鑑別 / / |
| (英) |
Hurst exponent / zero crossing rate / Parkinson’s disease / Major depression / vocal index / Identification / / |
| 文献情報 |
信学技報 |
| 資料番号 |
|
| 発行日 |
|
| ISSN |
|
| PDFダウンロード |
|