人工智慧數位語音助理系統開發平台
Development platform for AI Digital Voice Assistant Systems
效益與未來規劃
Benefits and Future Planning
｢人工智慧數位語音處理系統｣是由五個模組而形成，至終所發展的這五個模組將整合到單一晶片。首先是｢數位訊號處理｣模組，其的目的在於對輸入語音進行語音加強、語音分離等處理。
接著是｢語者辨識｣模組，它讓語音助理系統有能力辨識註冊語者的的聲音，只對註冊語者提供服務，而拒絕冒充者與欺騙攻擊等。
第三個模組是｢語音辨識｣，它的功用在是將使用者的語句轉換為文字。基於文字序列的資訊形態，｢自然語言處理｣模組就像人類的大腦一樣，能夠對文字序列進行語意分析（語意理解），進而提供適當的服務。
最後一個模組是｢語音合成｣，該模組能讓語音助理將所要表達的文字（所要傳達的意思）轉換為語音的資訊形態，提供更自然、更有效的互動方式。
The "artificial intelligence digital voice processing system" is formed by five modules, and the five modules developed will be integrated into a single chip. The first is the "digital signal processing" module, whose purpose is to perform processing such as Voice Enhancement and Voice Separation on the input voice.
Then there is the "Speaker Recognition" module, which enables the voice assistant system to Recognize the Voice of the Registered Speaker, and only provides services to the registered speaker, and rejects impersonators and deception attacks.
The third module is "Voice Recognition", its function is to Convert the User's Sentence into Text. Based on the information form of the text sequence, the "Natural Language Processing" module is like the human brain, which can perform Semantic Analysis (Semantic Understanding) of the text sequence and provide appropriate services.
The last module is "speech synthesis", which allows the voice assistant to convert the text to be expressed (the meaning to be conveyed) into a voice information form, providing a more natural and effective way of interaction.
使用課程
Course Used
數位訊號處理、語音辨識理論與實務、人工智慧、深度學習理論與實務
Digital Signal Processing, Speech Recognition Theory and Practice, Artificial Intelligence, Deep Learning Theory and Practice
高光譜影像相機
Hyperspectral Imaging Camera
效益與未來規劃
Benefits and Future Planning
農產品的品質問題是國內面臨的一大議題，目前農產品瑕疵篩選幾乎都是靠人工在分析缺陷方面，執行時間長，無法分析大量樣品且容易疲勞而誤判。
使用高光譜影像推掃式 VIS-NIR 高光譜影像相機取得農產品影像再運用人工智慧與深度學習量化、識別及分類瑕疵區域，搭配機器手臂進行自動化的整合，將有瑕疵的農產品在準確且 有效率的自動化方式進行篩選。
The quality of agricultural products is a major domestic issue. At present, the defect screening of agricultural products is almost always done by manual analysis of defects, which takes a long time to analyze a large number of samples and is prone to fatigue and misjudgment.
Use the hyperspectral image push-broom VIS-NIR hyperspectral image camera to obtain images of agricultural products, and then use artificial intelligence and deep learning to quantify, identify and classify defective areas. With the robot arm for automated integration, the defective agricultural products are screened in an accurate and efficient automated way.
使用課程
Course Used
圖形識別、高光譜影像處理技術與應用、巨量資料與圖形識別
Pattern Recognition, Hyperspectral Image Processing Technology and Application, Massive Data and Pattern Recognition
割草機機器人
Grass Mower Robot
效益與未來規劃
Benefits and Future Planning
以嵌入式平台，結合RTK-GPS、9軸感測器、卡爾曼濾波器、PID控制、超音波感測、影像辨識等技術，提供精準自主割草與避障能力之割草機器人。
研究重點：
－自走控制、避障：影像辨識、LiDAR、GPS
－路徑規劃：最短時間完成割草需求、最高之割草覆蓋率
－物聯網功能：主動告知何時充電、何時加油、何時保養、結合氣候狀況，預測何時需割草
－多台割草機協同合作
Based on the embedded platform, combined with RTK-GPS, 9-axis sensor, Kalman filter, PID control, ultrasonic sensing, image recognition and other technologies, it provides a mowing robot with precise autonomous mowing and obstacle avoidance capabilities.
Research Focus:
- Self-Propelled Control, Obstacle Avoidance: Image Recognition, LIDAR, GPS
- Path Planning: Complete the Mowing Demand in the Shortest Time, and the Highest Mowing Coverage Rate
- Internet of Things Function: Proactively Inform When to Charge, When to Refuel, when to Maintain, Combined with Weather Conditions, to Predict when to Mowing
- Collaboration of Multiple Lawn Mowers
頭戴式虛擬實境頭盔
Head-mounted virtual reality helmet
效益與未來規劃
Benefits and Future Planning
3D虛擬實境（VR）科技在教學上的應用越加廣泛，因具直覺性、多重感官刺激及遊戲互動式特性，讓使用者沉浸於虛擬建構的場域，提供身歷其境的體驗，較能吸引使用者注意。
藉由戴在頭上沉浸式頭盔及手持操縱桿虛擬操作堆高機，讓作業員熟練正確的叉取貨物程序，並檢視其操作是否符合規範，熟練標準程序。藉此虛擬實境輔助教育方式，讓員工事先學習和避免貨物倒塌的災害發生。
The application of 3D virtual reality (VR) technology in teaching is becoming more and more extensive. Because of its intuition, multi-sensory stimulation and interactive features of the game. It allows users to immerse themselves in the field of virtual construction, provides an immersive experience, and attracts users' attention.
By wearing an immersive helmet and a hand-held joystick to virtually operate the stacker, the operator is proficient in the correct fork and picking procedures, and checks whether the operation conforms to the specifications and is proficient in the standard procedures. With this virtual reality assisted education method, employees can learn in advance and avoid the disaster of cargo collapse.