◎系所教育目標：
為配合國家建設及產業發展之需要，本系以培育中高級資訊科技人才為目的。在教學理念上除了注重理論的探討之外並強調實際動手的能力，以期培育出具有深厚學識基礎並能實際應用的資訊科技人才。在專業必修中涵蓋基礎理論、電腦硬體、作業系統、資料結構及計算機網路等方面，並有畢業專題製作，使學生紮實基礎，同時課程包含四個專業學程，兼顧學術及實務之分流與訓練。分別為一：軟體工程及知識工程學程、二：互動多媒體學程、三：網路及資訊安全學程、四：資訊系統開發實務學程，以期作為日後升學就業的準備。 |
| ◎核心能力 | 關聯性 |
| 1.應用數理邏輯推理之能力 | 3 關聯性中等 |
| 2.發掘、分析及解決問題之能力 | 3 關聯性中等 |
| 3.現代資訊發展工具之使用與熟悉資訊應用系統之能力 | 2 關聯性稍弱 |
| 4.資訊軟體或硬體創新設計與實作之能力 | 2 關聯性稍弱 |
◎本學科內容概述：
In most multimedia applications, in order to reduce the volume of information to be transferred, a range of compression algorithms are applied to the different media types prior to integration them together. In addition to the compression algorithms that have been used for many years with text and images, there is now available a wide range of algorithms for the compression of speech, audio and video. Finally, the most modern emerging subjects, stereo photograph and stereoscope video processing technique, will be introduced for rapidly growing video-entertainment evolution. note : 教授encoding所用到各成分技術之內容時,會以機會學習方式,盡量將它們與發展自駕車所需之可能影像處理與深度學習網路中訊號處理機制做鏈結, 為增加新興智慧技術之了解與學習 |
◎本學科教學內容大綱：
1 Digital Signal Processing Concept including signal sampling and quantization 2 MPEG, JPEG introduction 3 Detail MPEG-4 4 Arithmetic coding and Huffman coding 5 Key differences introduction between MPEG 4 and H.264/HEVC 6 LZ77 text coding and its predictive concept 7 LZW dictionary-based text coding and its predictive concept 8 LMS-based regression processing to build a representative function for the use of interpolation, prediction and estimation. 9 Introduction of Investigating Kinect-based Stereoscopic video photographing and imaging; one view + one depth coding 10 Deep Learning Neural Network concept and principle |
◎本學科學習目標：
Brief illustration of audio-visual compression system curriculum for leave students to know below: And represented in a digital form, speech, audio and video, however, are generated in the form of continuously varying-normally referred to as analog signals. Hence in order to integrate all of the different media types together, it is necessary to first convert the various analog signals into a digital form. The integrated digital information stream can then be stored within a computer and transmitted over a network in a unified way. In addition, unlike text and images, which are created in the form of a single block of digital information, since speech, audio and video are continuously varying signals, the digitization process can produce large volumes of information which carries on increasing with time. Hence in most multimedia applications, in order to reduce the volume of information to be transferred, a range of compression algorithms are applied to the different media types prior to integration them together. In addition to the compression algorithms that have been used for many years with text and images, there is now available a wide range of algorithms for the compression of speech, audio and video. Until recently, however, because of the relatively low levels of compression that could be achieved, multimedia applications involving speech, audio and video-video-video telephony and video conferencing for example-required a high-capacity transmission channel to transmit the integrated source information. The rapid advances that have taken place in the field of compression over the past few years, however, mean that the capacity of the transmission channel required has been reduced to the point that most types of coders can now support a range of multimedia applications. Finally, the most modern emerging subjects, stereo photograph and stereoscope video processing technique, will be introduced for rapidly growing video-entertainment evolution.
Finally, I introduce the deep learning model by the relation of bases between deep learning model and compression.
教授encoding所用到各成分技術之內容時,會以機會學習方式,盡量將它們與發展自駕車所需之可能影像處理與深度學習網路中訊號處理機制做鏈結, 為增加新興智慧技術之了解與學習 |
| ◎教學進度： |
| 週次 | 主題 | 教學內容 | 教學方法 |
01
09/11 | Introduction | Digital Signal Processing Concept including signal sampling and quantization | 講授。 |
02
09/18 | MPEG | MPEG introduction | 講授。 |
03
09/25 | MPEG | Detail MPEG-4 | 講授。 |
04
10/02 | MPEG | Detail MPEG-4 | 講授。 |
05
10/09 | Lossless and entropy coding | Arithmetic coding and Huffman coding | 講授。 |
06
10/16 | predictive coding | DPCM and ADPCM | 講授。 |
07
10/23 | Common knowledge of DSP for video coding | Filtering principle and convolution | 講授。 |
08
10/30 | Text Coding | LZ77 text dictionary-based text coding and its predictive concept, | 講授。 |
09
11/06 | Text Coding | LZ77 text coding and its predictive concept | 講授。 |
10
11/13 | Middle Term Exam | Middle term examination I | 考試。 |
11
11/20 | Text Coding | LZW dictionary-based text coding and its predictive concept | 講授。 |
12
11/27 | Regression processing for deep leaning training: one practical tech. in numerical analysis | LMS-based regression processing to build a representative function for the use of interpolation, prediction and estimation for deep learning neural network (DNN) training. | 講授。 |
13
12/04 | 深度學習神經網路學習神經網路之基礎數理 | 深度學習神經網路學習神經網路之基礎數理 | 講授。 |
14
12/11 | 用於資料壓縮之深度學習神經網路 | 用於資料壓縮之深度學習神經網路 | 口頭報告。 |
15
12/18 | Middle Term Exam | Middle term examination II | 考試。 |
16
12/25 | 用於資料壓縮之深度學習神經網路 | 用於資料壓縮之深度學習神經 | 操作/實作、口頭報告、討論。 |
17
01/01 | 用於資料壓縮之深度學習神經網路困境與關鍵 | data compression for visual (AI) machine | 操作/實作、口頭報告、討論。 |
18
01/08 | Final Exam | Final examination | 口頭報告、考試。 |
◎課程要求：
有系統流程觀念 |
◎成績考核
課堂參與討論20% : 課堂參與討論為到課率
小考20%
期中考20%
期末考20%
書面報告20%
補充說明：到課率與上課態度記錄分數, 即算為"課堂參與討論"分數 |
◎參考書目與學習資源
1 用於資料壓縮之深度學習神經網路論文
2 自行開發教材 |
| ◎教材講義 請改以帳號登入校務系統選擇全校課程查詢方能查看教材講義 |