走出資料中心PECK 用類比晶片打造 AI與實體世界的互動平臺Beyond the Data Center: PECK Connects AI and the Physical World with Analog Chips
工商時報報導彭盛裕教授指導的跨校 PECK 團隊。團隊提出雙層類比閘控循環單元（AGRU）晶片，讓感測波形在類比領域直接進行時序運算，減少資料數位化與搬移的功耗及延遲；作品獲經濟部「智慧創新大賞」IC 設計類學生組銀牌。CTEE profiles PECK, a student team from NTUST and NYCU advised by Prof. Sheng-Yu Peng. Its dual-layer analog gated recurrent unit (AGRU) chip processes sensor signals directly in the analog domain to reduce conversion and data-movement overhead for edge AI. The project won Silver in the student IC design category of the MOEA Best AI Awards.
工商時報 CTEECTEE臺科大學生打造 AGRU 類比 AI 晶片 穿戴醫療續航力大躍進NTUST Students Develop an AGRU Analog AI Chip for Longer-Lasting Wearable Healthcare
彭盛裕教授團隊（PECK）開發 AGRU 類比 AI 晶片，直接在電流域處理類比訊號，免除 ADC 轉換與數位訊號處理，大幅提升穿戴醫療裝置的續航力與反應速度。神經脈衝分類準確率達 98%、心電圖分類 98.5%，整合感測功耗僅 99.36 µW；以 180nm 製程透過架構創新達成高效能，並相容 PyTorch 訓練框架。榮獲經濟部智慧創新大賞銀牌。Prof. Sheng-Yu Peng's team (PECK) developed an AGRU analog AI chip that processes analog signals directly in the current domain, eliminating ADC conversion and digital signal processing to improve the battery life and response time of wearable medical devices. The chip achieves 98% neural-spike classification accuracy and 98.5% ECG classification accuracy with total sensing power of only 99.36 µW. Fabricated in a 180 nm process, it delivers high efficiency through architectural innovation, remains compatible with the PyTorch training framework, and received a Silver Award at the MOEA Best AI Awards.
工商時報 CTEECTEE