遠端語音監控製造系統

現今智慧語音助理的應用無所不在，許多消費性產品不僅能聲控還能進行對話，譬如：智慧型手機、智慧音箱、智慧家電…等，以最直覺的方式進行聲控，但將智慧語音助理與智慧工廠結合的案例卻鮮少，最初成功的案例源自於美國iTSpeeX 公司於2018 年推出語音助理雅典娜（Athena），主要使用降噪耳機麥克風下達語音指令控制CNC 機台，藉以降低技師操作技術門檻及人事成本，但侷限於現場進行聲控，無法遠端語音監控。2019 年12 月全球爆發嚴重特殊傳染性肺炎疫情(COVID-19)，為避免群聚感染，工廠採用分流上班制甚至停工，使經濟受到嚴重的影響，因此本研究著重於以遠端下達語音指令監控各工廠的製造設備，不僅能減少群聚的危險，而語音助理更是操作員的小幫手。隨著物聯網的興起進而讓智能製造逐漸崛起，透過遠端語音監控製造設備並蒐集相關之數據進行決策分析，已成為不可或缺的重要因素。綜合以上所述，本研究建置「遠端語音監控製造系統」，並細分為以下三項子系統：(1)「遠端語音控制製造設備」子系統：藉由APP 或於ReSpeaker Mic Array 遠端下達語音指令對各工廠之不同設備進行控制；(2)「數據異常通知」子系統：當製造設備之數據發生異常時，透過通訊軟體（如：Line、簡訊、E-mail）接收異常訊息；(3) 「數據圖表及監視影像」子系統：管理者可於瀏覽器中查看各工廠的視覺化動態圖表數據與監視影像，以即時掌握各工廠之設備狀態。

關鍵字

語音辨識； Google 雲端語音服務；階層式jieba 字典

並列摘要

Nowadays, the application of smart voice assistant is ubiquitous. Many consumer products can not only control voice, but also conduct dialogue, such as smart mobile phones, smart speakers, smart home appliances, etc. they control voice in the most intuitive way, but few cases combine smart voice assistant with smart factory, The first successful case comes from the launch of Athena, a voice assistant by an American itspeex company in 2018, which mainly uses noise reduction earphone microphone to issue voice commands to control CNC machine, so as to reduce the technical and personnel costs of technicians, but limited to onsite voice control. In December 2019, the global outbreak of severe special infectious pneumonia (covid-19) broke out. In order to avoid cluster infection, some factories adopted the system of separate work, and even shut down, which seriously affected the economy. Therefore, this study focused on the voice command from the remote end to monitor the manufacturing equipment of each factory, which could not only reduce the risk of cluster, The voice assistant is the operator's little helper. With the rise of the Internet of things, intelligent manufacturing gradually rises. It has become an indispensable factor to control manufacturing equipment through the network and collect relevant data for decision analysis. Based on the above, a remote voice monitoring manufacturing system is built and divided into the following three subsystems: (1) the remote voice control manufacturing equipment subsystem: controlling different equipment in each factory by voice commands from app or remote end of the speaker mic array. (2)"Data exception notification" subsystem: when the data of manufacturing equipment is abnormal, receive the exception message through the communication software (such as line, SMS, e-mail). (3) "Data chart and monitoring image" subsystem: managers can view the visual dynamic chart data and monitoring image of each factory in the browser, so as to grasp the equipment status of each factory in real time.

並列關鍵字

speech recognition ； Google Cloud Voice Service ； Hierarchical jieba dictionary

參考文獻

[1]Viterbi, A. (1967). Error bounds for convolutional codes and an asymptotically optimum decoding algorithm. IEEE transactions on Information Theory, 13(2), 260-269.

Google Scholar

[2]Williamson, B. (2016). Digital education governance: data visualization, predictive analytics, and ‘real-time’policy instruments. Journal of Education Policy, 31(2), 123-141.