構音異常溝通輔具之人工智慧系統與晶片
Auditory-Cognition AI System and System-on-Chip Designs for Dysarthria
計畫團隊成員 Members
王進賢特聘教授
Distinguished Prof.
Jinn-Shyan Wang
國立中正大學電機系&晶研中心
EE Dept. & SoC Research Center, CCU
賴穎暉副教授
Assoc. Prof.
Ying-Hui Lai
國立陽明交通大學生醫系&
晶研中心
BME Dept., NYCU & SoC Research Center
葉經緯教授
Prof. Ching-Wei Yeh
國立中正大學電機系&晶研中心
EE Dept. & SoC Research Center, CCU
林泰吉副教授
Assoc. Prof.
Tay-Jyi Lin
國立中正大學資工系&晶研中心
CSIE Dept. & SoC Research Center, CCU
Category
Dysarthria, Artificial Intelligence, Deep Learning, Low Power, System on a Chip
技術亮點
Technical Highlights
技術與產品現況
- 未有(構音異常)口譯輔具及概念。
- 多數產品著重在電腦字卡輔助患者教育/復健訓練。
成果亮點-關鍵技術
- 首創並展示中風與腦麻病患構音異常聲音轉換先例。
- 提升VC效益，並降低90%病人訓練負擔。
- 進化DNN-Based VC技術為構音異常病患之隨身溝通輔具技術。
- 建構單一功能隨身溝通輔具雛形，以成為健保給付之可能或可行方案。
- 低功耗DNN設計+低漏電近鄰界電壓電路技術，降低>80%功耗。
- Dysarthria is a common problem of patients with neurological diseases, such as stroke or the Parkinson’s disease, which affects their quality of life.
- The 4-year project “Auditory-Cognition AI System and System-on-Chip Designs for Dysarthria” will develop AI-based dysarthria-voice-conversion, low-power embedded system, and system-on-a-chip design technologies for achieving the goal of “Speak for Dysarthria.”
- In the first-year project term, a pure-software-based Dysarthria Voice Conversion (DVC) prototype device running on a 667-MHz-dual-Cortex-A9-Zynq platform has been designed and constructed, which can perform inside tests of converting a dysarthria voice into an intelligible one in less than 1 second.