1270831 九、發明說明： 【發明所屬之技術領域】 位風ί發明係有關於—種具備數位典藏機制的多媒體數 予白糸統，尤其是指-種更有效及流暢的視訊典藏技 可將老師上課所用到的投影片或是手寫板上所寫的資 旦^乂及課堂上可能發生的—些對話過程和學生跟老師的 Γ =以處理，以便日後方便存取之具備數位典藏機制的 夕螺體數位學習系統創新設計者。 【先前技術】 按’目前市面上有許多即時視訊軟體，例如J〇inNet、 Cen<trac和NetMeet i ng等，雖然有些視訊軟體甚至可同時 支^ 16個人的晝面，但是大部分軟體並未提供能將當時 匱土兄併儲存之技術。例如j〇inNet，提供了許多線上教 予的功此·互動式教學板，提供老師可以手寫輸入並廣播 ，所有人的功能；控制面板（包含人數統計、線上投票 等）；即時討論群組等。但是目前的J〇inNet在影音資料 儲存的過程中，卻只有將老師的上課畫面錄製下來，對於 之後的使用者而言，也只能撥放老師上課的晝面，並未能 呈現線上教學時老師與學生互動的景象，在影音資料儲存 的搜尋擷取過程中，目前僅限於利用章節來分段，再由使 用者透過超鏈結（Hyper Link)來將各章節做連結，以提供 點選撥放’卻無法達到快速找到欲學習的主題。 即時影音技術早已被廣泛地使用在 E-learning 及視 訊會4系統上；E—learning及視訊會議其實是一體兩面的 問題’共用的技術也差不多相同，雖然許多的商用軟體也 都號稱有即時錄影的功能，不過其皆並無法像當時的狀況 1270831 重現，所謂的即時錄影，只不過是將發話者的影像與聲音 c錄下來而已，並沒有記錄時間順序，如此可能導致事後 無法重造當時會議狀況；另一個可能引發之問題是無法提 供搜尋之功能，這會讓欲使用會議記錄或教學記錄的人變 得沒有效率，因為使用者必需重頭看到完才能找到想要看 的資料。 a刖的即時視訊的錄製並不理想，大多都不支援視訊 ，影的功能，對於即時視訊的儲存有支援的J〇inNet來 说’由於其在錄製的過程中會以每i小時的影片壓成1〇肫 的檔案壓縮率對影像作破壞性的壓縮，使得影像的品質嚴 f的失真、頓格且晝面變的不連續，對於一般的使用者而 。，面對影像強烈的失真及不連續，往往造成使用率的下 降，再加上一般的即時視訊都只是單純的撥放錄製主持人 與參與者的影像及聲音，並未將其他的上課教材及一些投 影片-併播放，而且是單-使用麵方式誠無法表達時 間重疊的效果。 【發明内容】 本發明之具備數位典藏機制的多媒體數位學習系 統，其主要係發展一個更有效及流暢的視訊典藏技術，將 老師上課所制的投影片或是手寫板上所寫的資料， 課堂上可能發生的-些對話過程和學生跟老師的 以處理以便日後枝存取，換句話說，n祕夠將 上課教材及視訊影像一併整合儲存，並利用XML半結 的語言及時間軸控制來進行控制，當使用者欲 ^ 訊時，藉由XML格式的log fiie來读跑此士 ，、減視 不靖取所有相關性的警 料’這些資料包含：在不同影料服_影音資料 1270831 投影片以及Html資料，以及在適當的時間點所呈現的影 音資料，透過本典藏技術的應用，使得在撥放的過程中能 夠呈現原本的上課狀態，同時也利用文字來對視訊進行搜 尋，使用者只需輸入關鍵字，伺服器則會自動比對所有的 影音賀料’並顯示目剷的影音資料在某一時間段落有相關 的主題，並可直接跳躍到該時間點播放，不需從頭循序撥 放，以快速提供使用者取得其感興趣的資料。 【實施方式】 首先，本發明系統實施方式共有系統資料結構雛形、 時間軸之事件S己錄、時間軸、計時器與基因演算法等；該 系統資料結構雛形提供研究人員在實作時所需要用到的 檔案、要用什麼檔案才能把情境清楚地儲存在系統中；時 間軸之事件記錄採用XML格式，目的希望系統能提供盥装 它機器設備溝通的能力；計時器中記錄了某個時點^事 件，時間軸則是提供了同一時間内可以同時執行不同動作 的能力，例如··在某個時間點上剛好有二個使用者在發言 及一人在電子白板上晝了重點而產生筆跡，這三個動作必 需一起呈現給使用者看，這些是由時間轴與計時器一起運 作來達成的。 系統資料結構離形：1270831 IX. Description of the invention: [Technical field of invention] The invention of the invention relates to a kind of multimedia number with a digital collection mechanism to the Cretaceous, especially to a more effective and smooth video collection technique. The slides used are either the Zidan 乂 written on the tablet and the dialogue process that may occur in the class and the 学生 学生 学生 学生 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以 以Innovative designer of the digital learning system. [Prior Art] According to 'there are many real-time video software on the market, such as J〇inNet, Cen<trac and NetMeet i ng, although some video software can even support 16 people at the same time, most software does not. Provide the technology to store and store the brothers at that time. For example, j〇inNet provides a lot of online teaching and interactive teaching boards, providing teachers with handwriting input and broadcasting, all of them's functions; control panel (including number of people, online voting, etc.); instant discussion groups, etc. . However, in the current process of storing audio and video data, J〇inNet only records the teacher's class screen. For the subsequent users, it can only be used for the teacher's class and fails to present online teaching. The interaction between the teacher and the student is limited to the use of chapters for segmentation in the search and retrieval process of audio and video data storage. The user then links the chapters through hyperlinks to provide points. Dialing 'but can't reach the topic of quick learning. Instant video and audio technology has been widely used in E-learning and videoconferencing systems; E-learning and video conferencing are actually two-in-one issues. The shared technology is similar, although many commercial software also claims to have instant video. The function, but it can not be reproduced like the situation 1270831 at that time. The so-called instant recording is just recording the video and sound of the speaker, and there is no chronological order, which may lead to the re-creation of the time. The status of the meeting; another problem that may arise is the inability to provide a search function, which makes the person who wants to use the meeting record or the teaching record inefficient, because the user has to see it again before he can find the information he wants to see. A刖's instant video recording is not ideal. Most of them do not support video and video functions. For J〇inNet, which supports instant video storage, 'because it will be pressed every hour during the recording process. The file compression rate of 1〇肫 is destructively compressed on the image, so that the quality of the image is severely distorted, and the surface of the image is discontinuous for the average user. Faced with the strong distortion and discontinuity of the image, the usage rate is often reduced. In addition, the general instant video is simply the video and sound of the recording host and the participants, and no other teaching materials and Some slides - and play, and single-use face mode can not express the effect of time overlap. SUMMARY OF THE INVENTION The multimedia digital learning system with the digital collection mechanism of the present invention mainly develops a more effective and smooth video collection technology, and the materials written by the teacher on the slides or on the tablet, classroom Some of the dialogue processes that may occur and the processing of the students and the teachers for future access, in other words, n secrets to integrate the textbooks and video images together, and use the language and timeline control of the XML half-junction To control, when the user wants to control, the log fiie in XML format is used to read and run the sergeant, and the refusal to take all the relevant warnings. 'These materials include: in different shadow clothing _ audio and video materials 1270831 The film and Html data, as well as the audio and video materials presented at the appropriate time, through the application of this collection technology, can make the original class state in the process of playing, and also use the text to search the video. The user only needs to input the keyword, the server will automatically compare all the video and audio gifts 'and display the video data of the eye shovel at a certain time. Fall on related topics, and can jump directly to that point in time to play, you do not need to put aside sequentially from the beginning, to provide users with fast access to information which is of interest. [Embodiment] First, the system embodiment of the present invention shares the prototype of the system data structure, the event of the time axis, the time axis, the timer, the gene algorithm, and the like; the prototype of the system data structure provides the researcher with the need for the implementation. The files used, the files to be used to store the situation clearly in the system; the event record of the timeline is in XML format, the purpose is to hope that the system can provide the ability to pretend its machine equipment communication; the timer records a certain point in time ^Events, the timeline provides the ability to perform different actions at the same time, for example, at a certain point in time, just two users are speaking and one person is focused on the electronic whiteboard to generate handwriting. These three actions must be presented to the user together, which is achieved by the timeline working with the timer. System data structure is out of shape:
請參閱第一圖所示，其係為一個真實教學環境到資料 結構的演化過程’使用者的影像與聲音可以使用RTP stream來建構，投影片可以採用Microsoft Power Point 來製作，另外使用者的行為則將它定義為系統中之事件， 最後以 video、audio、gif、html 及 log file 來儲存。 時間軸之事件記錄： 7 1270831 參閱第二圖所示，採用XML之格式來記錄，此事件記 錄包括群組中任一個參與者的任何一個動作，例如按下 “我要發言”的按鈕，或者是伺服器的一連串動作，例如 伺服器收到某個令命“錄影開始” or “換投影片，，。 時間軸： / 參閱第三圖所示，Y軸為檔案項目，X軸為時間序列， 在X軸下面表示投影片與記錄檔所顯示的時間，而上方 ABCDE表示影音資料播放的時間，Screen與Screen之間 _ 的垂直虛線表示投影片之切換點，於第三圖中顯示出投影 片切換時會切割到影音資料檔，例如A與D影片會存成兩 個部分，B與C影片會存成三個部分，E會存成五個部分， 以Screenl而言有Teacher及User 1正在播放中；對 Screen 2來說有Teacher、Userl及User2正在播放中； Screen 3 有 Teacher、User 2 及 User 1 正在播放中；Screen 4有Teacher、User 2及User 1正在播放中，但是[Jser2 分成兩個部分並不一定是同一個參與者，User 2僅代表一 φ 個即日守影音通吼頻道；最後對Screen 5而言有Teacher、 User 1及User 2正在播放中，於第三圖中有兩份Useie i 及User 2檔及一分Teacher檔，這些都只是代表即時視 sfl的通訊頻道，除了 Teacher是由固定的主持人使用外， 其它則可能由參與者的任何一人來使用，所以第三圖中的 兩& User 1可能代表著不同的學生，每個檔案的開始與 結束均會在log file中留下時間記號，一但被計時器觸 發就會進入播放程序。 以GA求RTP/RTCP之最佳參數組合： 1·已知某一 AV server之實體線路頻寬：χ 1270831 2· Video 之參數集合為{encoding，frame，size} a.encoding={JPEG ， H263} b·frame={l ， 5 ， 10 ， 12 ， 25 ， 30} c.size={160*120 ， 176*144 ， 320*240} 3. audio 之 encoding={Linear ， G723} 4. 每一個擁有發言權之User所需要的頻寬=仄Please refer to the first figure, which is a real teaching environment to the evolution of the data structure. The user's image and sound can be constructed using RTP stream. The slide can be produced by Microsoft Power Point, and the user's behavior. It is defined as an event in the system, and finally stored as video, audio, gif, html, and log file. Timeline event record: 7 1270831 See the second figure, recorded in XML format, this event record includes any action of any participant in the group, such as pressing the "I want to speak" button, or It is a series of actions of the server. For example, the server receives a certain command "record start" or "change slide." Timeline: / See the third figure, the Y axis is the file item, and the X axis is the time series. The time below the X-axis indicates the time displayed by the slide and the log file, and the upper ABCDE indicates the time when the video material is played. The vertical dotted line between Screen and Screen indicates the switching point of the slide, and the projection is displayed in the third figure. When the film is switched, it will be cut into the audio and video data files. For example, the A and D movies will be saved in two parts, the B and C videos will be saved in three parts, and the E will be saved in five parts. In the case of Screenl, there are Teacher and User 1 Playing now; for Screen 2, Teacher, Userl, and User2 are playing; Screen 3 has Teacher, User 2, and User 1 playing; Screen 4 has Teacher, User 2, and User 1 playing, Yes [Jser2 is divided into two parts and not necessarily the same participant. User 2 only represents one φ day-to-day video channel; in the end, for Screen 5, Teacher, User 1 and User 2 are playing, in the first In the three pictures, there are two Useie i and User 2 files and one point Teacher file. These are just the communication channels representing the instant view sfl. Except that the Teacher is used by a fixed moderator, others may come by any one of the participants. Use, so the two & User 1 in the third picture may represent different students, the beginning and end of each file will leave a time mark in the log file, once triggered by the timer will enter the player. The optimal parameter combination of RTP/RTCP is obtained by GA: 1. The physical line bandwidth of an AV server is known: χ 1270831 2· The parameter set of Video is {encoding, frame, size} a.encoding={JPEG , H263 } b·frame={l , 5 , 10 , 12 , 25 , 30} c.size={160*120 , 176*144 , 320*240} 3. audio encoding={Linear , G723} 4. Each The bandwidth required by the user who has the right to speak =仄
Ui{(encoding y frame，size)， (encoding) }氺（person+1) 5. AV server必需提供之最小總頻寬：7，即適配函數。 priority=l，Ui定義為Ai，m代表數量，代表身分是講 師。 priority=0，Ui定義為Bj，k代表數量，代表身分是學 生。 n=m+k ° n m k 6·最4化#件ί : /3Ui{(encoding y frame,size), (encoding) }氺(person+1) 5. The minimum total bandwidth that the AV server must provide: 7, the adaptation function. Priority=l, Ui is defined as Ai, m is the number, and the representative is the lecturer. Priority=0, Ui is defined as Bj, k represents the quantity, and the identity is the student. n=m+k ° n m k 6·最四化#件ί : /3
X>T>0.SXX>T>0.SX
=>1.25 >-^—>1 0.8X => 1.25 > β>\ 召如愈接近1，我們定義為愈好，但不可以ΘΜ.25，這表 示超出AV server實體線路頻寬之負荷，若都不在 1.25〜1之間我們可以允許次佳解ι>θ。 7.最佳化條件2 : α 要符合 distance = teacher - student =正數 1270831 3 - Β』=正數：^ q 8·本發明以7個bit來表示個別數組合之型態 {yideo—encoding > frame frame frame , size size , audio—encoding) 右某-個空間值沒有代表意義，則重新亂數分配。 9.實作方法 a·隨機為仿選出7個bit，前观個分配給Ai，其它的配 給Bj=>1.25 >-^—>1 0.8X => 1.25 >β>\ Calling closer to 1, we define the better, but not ΘΜ.25, which means that the AV server physical line frequency is exceeded. The load of the width, if not between 1.25 and 1 we can allow the second best solution ι > θ. 7. Optimization condition 2: α to meet distance = teacher - student = positive number 1270831 3 - Β 』 = positive number: ^ q 8 · The present invention uses 7 bits to represent the type of individual number combination {yideo-encoding > Frame frame frame , size size , audio—encoding) If a space value of the right does not represent meaning, then the random number is allocated. 9. Implementation method a· Randomly select 7 bits, copy the front view to Ai, the other match Bj
b·疋義母一種encoding與size之組合，以每秒一個b·疋义母 A combination of encoding and size, one per second
frame為基楚，計算出頻寬。 我們定義年如工^^元夕二_唓列 JPEG， 320*240 JPEG， 176*144 JPEG， 160*120 H263， 320*240 H263， 176*144 H263， 160*120 a·定義各別的person數量 b·計算出7，如下所示： (encode一 size[encoding][size] * frame[frame_nwnber]+audio) * (person+1) c·代入々，若發生/3>1·25，表示它不適合生存，則會 殺掉這條基因。 d.記錄現在所有的Ai與艮值 e·父配後給予突變’突變率為〇· 5 ’針對f 與gize 來突變，若1 >石則不交配直接突變，回到步驟d，直 到交配}^代之後或找出P個最佳的值就停止。 10 1270831 f · k擇s己錄中最好的參數組合。 10·基因交配方式 是以instance為主體，它代表一組AV server的 ，數組合，其内容為每一個User所要使用的參數，本 研九的基因父換方式如第四圖所示是ins^ance内的基 因作局部交換的動作。 計時器： 清參閱第五圖所示，垂直的linked list表示時間 • 每1時間點均有可能有衫事件或是沒有事件，一 十夺器執行至該時間點，即會將它所承載的事件分別交 由獨立的thread去執行。 影像搜尋索引技術： 少凊參閱第六圖所示，其顯示座標中γ軸的值愈小表示 在〜片中愈後面的片段，故可以推出ρρτι的¥值比ρρτ ^，Υ值大，X軸座標標愈小表示索引詞的重要性愈大也 影音串流之内容意義愈相關，實際上只依ρρτ和文 • 纟交談内容來做索引且令ρρτ之相關性大於文字交談内容 之相關性，由於語音辨識技術尚未成熟，若要辨識都需要 、、二過曰頻的訓練，且辨識正確率也不高，因此本發明不採 ^扣曰辨識，也就是說搜尋的關鍵字是存在於ppT與文字 乂淡中的，若有相關詞則會找出其對應的影音資料，如第 六圖所示在同一 ΡΡΤ中影音資料可能是重疊出現的， 必需要能一起播出來。 、、 曰請再參閱第七圖所示，本發明欲開發的系統最終目的 是要可以將真實情境透過資訊科技技術來傳遞與儲存，儲 存之後也要能夠提供搜尋的功能，這可以讓現代忙錄的人 11 1270831 效二:學習，甚至也可以讓身體殘障人士獲得學 部分，各別說明如^過^成即時學習與非即時學習兩個 (Α)即時參與上課： 便會C，二以」教=:二-但使用者登入後系統 限，而學生’老師角色擁有較多的權 b乖乖聽課」不能亂說話，除非老師開 3限-戶!：老師擁有主持課堂秩序的權限，整個課程結 束後二糸統：自動執行「整合異地播案」的工作，整合完 成後就成功地結束本次上課過程。 (B)觀看已上過的課程： -般人在上課時多少都會做些筆記，但筆記可能也會 ί=Γ=以如果能回到上課的時間，那麼問題就 因此本研究在儲存上課資料方面加上了 狀：二1:太C又计當然使用者也可以自己選擇章 非線性學習上最大的特色是可以查詢 奎、：合k方便讓使用者不必從頭看到完才能找；答 案’糸射依照制麵人㈣鍵字去 段’這個概念類似像入口網站的搜尋引擎，影片== 鍵字的關係？由系統自動對應’因此可以降低人工成本。 在整個課㈣會議結束後自動立刻崎檔案整合的 動作’確保不會因為一個網路節點失效，而全部皆：。 我們可以令VM提供Data center服務，—但有某個課程 或會議結束VM會自動進行記錄㈣析，接著將所有RTp 貧料從各AV server中取回並轉權，由packet⑽此檔 12 1270831 隻成AVI檔，包括Player —併儲存在同一目錄下，放便 製作成光碟形式。 藉由以上所述，本發明主要是針對目前用在 LLearning或視訊會議的即時視訊軟體及v〇D的串流伺 服器之缺點加以改善，並結合同步即時及非同步即時的優 點’以提升即時視訊系統的品質；其具有下列優點： 1·本發明利用加入文字搜尋的功能，讓使用者依據自己所 想了解的議題做搜尋，可以直接跳要串流視訊的某一個 時間點，並從中間開始聽起，而不需每次都要從一開始 聽起。 2·本發明利用基因演算法找AV server中所能負荷的網 路負載量，更進一步求出每一個User的RTP最佳參 數，目的是期望AV server不會因為無法負荷所需的 Multicasting封包而導致RTP遺失，並且在使用人數少 時不會造成網路產能的浪費，此時也可以提高影像與聲 音的品質。 3·本發明由於在錄製的過程中，會同時將所有人的影音錄 製到不同的影音伺服器，當要再播放時會透過一個「同 步時間轴控制」的機制，將不同的影音伺服器的影音資 料做同步播放，並將配合所使用到的文字和圖片等教學 教材也一並完整呈現。 4. 本發明是以Java開發，而Java Applet是可以將 application嵌入在網頁中的一種技術，對客戶而言他 們只會看到網頁’所以只需劇覽器即可操作本系統讓使 用者達到真正零安裝，以及隨時隨地使用的效果。 5. 本發明由於在串流視訊的錄製過程中，可能同時由多台 13 .1270831 景々音伺服器分散式的錄製，對於何時發生相關事件（例 如：發言權的轉換）並透過l〇gfile來紀錄成XML格式， 以便未來方便讀取影音串流資料。而XML具有自我定義 標籤的特性，並透過XMLSchemal來描述XML格式的log file，使的將來可做資料交換之用途。 6·本發明結合了線性可以同時進行教學互動較佳，及非線 性則不需要大家在共同的時間上線學習，隨時隨地就可 以學習，兩者的優點利用GA讓線性可以具有q〇s的特 性’將大家共同學習的過程情境完整且流暢地錄製下 來，供未來複習。 7·本發明中的投影片可以支援Power Point，此投影片正 是教學内容搜尋的關鍵，因此演講者自己的内容可以被 容易搜尋出來，減少搜尋的錯誤率，增加教學内容與搜 哥的吻合率。 綜上所述，本發明實施例確能達到所預期之使用功 政’又其所揭露之具體構造，不僅未曾見諸於同類產品 中’亦未曾公開於申請前，誠已完全符合專利法之規定與 要求’爰依法提出發明專利之申請，懇請惠予審查，並賜 准專利，則實感德便。 14 .1270831 【圖式簡單說明】 - 第一圖：本發明之資料結構儲存形式圖 . 第二圖··本發明之XML格式圖 第三圖：本發明之時間轴示意圖 第四圖：本發明之基因交配方式圖 第五圖：本發明之計時器示意圖 第六圖：本發明之搜尋機制概念圖 第七圖：本發明之系統運作流程圖 15Frame is the base, and the bandwidth is calculated. We define the year as ^^元夕二_唓 JPEG, 320*240 JPEG, 176*144 JPEG, 160*120 H263, 320*240 H263, 176*144 H263, 160*120 a·Define individual persons The number b· is calculated as 7 as follows: (encode one size[encoding][size] * frame[frame_nwnber]+audio) * (person+1) c·substitute 々, if /3>1·25, It is not suitable for survival, it will kill this gene. d. Record all current Ai and 艮 value e·parent after giving mutation 'mutation rate 〇· 5 ' for f and gize to mutate, if 1 > stone does not mate directly mutation, return to step d until mating }^ After the generation or find the P best values, stop. 10 1270831 f · k Choose the best combination of parameters in s recording. 10. The gene mating method is based on instance, which represents a set of AV server, number combination, and its content is the parameter to be used by each User. The research method of the gene parent of this research is shown in the fourth figure is ins^ The genes in ance are used for local exchange. Timer: As shown in the fifth figure, the vertical linked list indicates the time. • There may be a shirt event or no event every 1 time. When the timer is executed until that time, it will carry it. The events are handled by separate threads. Image search indexing technology: As shown in the sixth figure, the smaller the value of the γ-axis in the coordinate is, the smaller the value is in the slice. Therefore, the value of ρρτι can be derived as ρρτ ^, which is larger, X The smaller the axis coordinates, the greater the importance of the index words and the more relevant the meaning of the video stream. In fact, only the ρρτ and Wen• 纟 conversations are indexed and the correlation of ρρτ is greater than the relevance of the text conversation. Since the speech recognition technology is not yet mature, if it is necessary to identify the training, and the recognition rate is not high, the present invention does not use the identification, that is, the searched keyword is present in the ppT. If there is a related word, it will find the corresponding audio and video data. As shown in the sixth picture, the audio and video data may overlap in the same video, and it must be broadcast together. Please refer to the seventh figure. The ultimate goal of the system to be developed by the present invention is to transfer and store the real situation through information technology. After storage, it can also provide search function, which can make the modern busy. Recorded person 11 1270831 Effect 2: Learning, you can even let the physically handicapped people get the part of the school, each description such as ^ ^ ^ into instant learning and non-instant learning two (Α) instant participation in class: will be C, two to "Teach =: two - but the user is limited after the system is logged in, and the student 'teacher role has more rights b 乖乖 class" can not talk nonsense, unless the teacher opens a limit of 3 - household!: The teacher has the authority to preside the classroom order, the whole At the end of the course, the two systems: automate the work of “integrating the remote site broadcast”, and successfully complete the course after the integration. (B) Watch the courses that have already been taken: - Most people will take some notes during class, but the notes may also be ί=Γ=If you can return to class time, then the problem is therefore stored in the study. Aspects added: 2: too C and count, of course, users can also choose their own chapters. The biggest feature of nonlinear learning is that you can query Kui, which is convenient for users to find without having to look through the head; According to the face-to-face (four) key word segmentation, the concept is similar to the search engine of the portal site, the relationship of the film == key word? Automatically responded by the system' thus reduces labor costs. Immediately after the end of the class (4), the action of automatically syndicating the files is ensured that no one network node will fail, but all: We can make the VM provide the Data center service—but there will be a record or conference end VM will automatically record (4), then all RTp poor materials will be retrieved from each AV server and transferred to the right, by packet(10) this file 12 1270831 only Into the AVI file, including the Player - and stored in the same directory, and put it into a disc format. In view of the above, the present invention is mainly directed to improving the shortcomings of the current video software and v〇D streaming server used in LLearning or video conferencing, and combining the advantages of synchronous instant and asynchronous instant to improve instant. The quality of the video system; it has the following advantages: 1. The invention utilizes the function of adding a text search, allowing the user to search according to the topic that he wants to know, and can directly jump to a certain time point of the streaming video, and from the middle Start listening, not every time you have to listen to it from the beginning. 2. The present invention uses a gene algorithm to find the network load that can be loaded in the AV server, and further finds the RTP optimal parameter of each User, in order to expect that the AV server will not be unable to load the required Multicasting packet. This leads to the loss of RTP and does not waste network capacity when the number of users is small. At this time, the quality of images and sounds can also be improved. 3. The present invention records the audio and video of all people to different audio and video servers at the same time during the recording process, and when playing back, the different audio and video servers will be transmitted through a "synchronous time axis control" mechanism. The audio-visual materials will be played synchronously, and the teaching materials such as texts and pictures used will be fully presented. 4. The present invention is developed in Java, and the Java Applet is a technology that can embed an application in a web page. For the customer, they only see the web page' so that only the browser can operate the system for the user to reach True zero installation, and the effect of using it anytime, anywhere. 5. The present invention may be distributed by multiple 13.1270831 landscape voice servers during the recording of streaming video, for when related events occur (for example: conversion of the floor) and through l〇gfile To record in XML format, in order to facilitate the reading of video streaming data in the future. XML has the characteristics of self-defined tags, and XMLSchemal is used to describe the log file in XML format, so that it can be used for data exchange in the future. 6. The invention combines linearity to facilitate simultaneous teaching interaction, and non-linearity does not require everyone to learn online at a common time, and can learn at any time and place. The advantages of both use GA to make linearity have the characteristics of q〇s. 'The process of learning the process of learning together is complete and smooth for future review. 7. The slide in the present invention can support Power Point, which is the key to teaching content search, so the speaker's own content can be easily searched, reducing the search error rate and increasing the teaching content to match the search. rate. In summary, the embodiments of the present invention can achieve the intended use of the power policy and the specific structure disclosed therein, which has not been seen in the same product and has not been disclosed before the application, and has fully complied with the patent law. The regulations and requirements ' 申请 apply for invention patents according to law, please apply for review, and grant patents, it is really sensible. 14 .1270831 [Simple description of the drawings] - First figure: storage structure diagram of the data structure of the present invention. Second diagram · XML format diagram of the present invention Third diagram: Schematic diagram of the time axis of the present invention Fourth diagram: The present invention Figure 5: Schematic diagram of the timer of the present invention: Figure 6: Conceptual diagram of the search mechanism of the present invention: Figure 7: Flow chart of the operation of the system of the present invention 15