M328569 八、新型說明: 【新型所屬之技術領域】 本創作係關於一多媒體裝置,其特別係關於—種提供互動式數位 影音多媒體服務之多媒體鏡。 【先前技術】 隨著科技的日新月異,發展落實數位家庭、營造智慧生活 ϋ市場趨勢,而在數位家電的應用中,數位影音裝置$逐漸趨j :力^ ’而在顯不||的領域中,已具有鏡面及顯示功能之相關專利, i lilt aT〇7563 /Αν"1#ϋ 晶顯示器’而同時用於顯示目的之鏡模組 = ,自顯μ的先傳輸,並可確實將來自螢幕區外部的 曰: =所知用的極化鏡’幾乎可讓光去穿透鏡面,因而可使綱示哭ΐί 力能’使該鏡模組於使用狀態下為—顯示器,而於 狀心下僅為-面鏡子;但該極化鏡的_ ^使用 般大眾’再者該鏡模組僅提供面及普及 無法提供-絲式醜位影音多舰服:成扣料向輸出影音, 【新型内容】 II於以上所知,本創作提供—種 =多媒體服務,該多媒體鏡包括-鏡面單互動式數位 農置、—影音辨識處理裝取裝置、—音訊錄 i連結關係是藉崎影音辨識處理裝音訊輪出震置, 、该音觸轉置、郷雜彡像梅取裝 其中,該鏡面單元前後兩面的光學性質相同之特性,當該鏡面單 5 M328569 元前後兩面的光度環境不同時,光線強度越強之一 反之’光線強度微弱之-面,雖會有部份光線穿透該鏡、,了门 光度微弱,使縣線強度較強之—面將I法看 ①f70 ’但因 創作之多媒體鏡即藉由此原理執行,之影像,而本 會高於該多媒體鏡外部的光職度,㈣達成==株源強度又 體鏡即可啊具魏㈣能與顯示魏。、…' ’因此該多媒 而該影像擷取裝置、該音訊擷取裝置、該 輸出裝置係分別電氣連結於該影音辨識處理心置與該音訊 裝置可為電_合树(awge C()upled Devi;e,’綠擷取 置’係用以擷取-影像訊號,而該音訊擷取裝置^機等裝 擷取裝置,將·取㈣取^與該音訊 進行比對、辨識,或該資料庫亦謂存辨科庫可提供資料 或该音訊輪出 ^辨識處理裝置處理後之訊號,將藉由該^輪二=;而經由影 i置回饋一影像或音訊訊穿 冢輸出衣置或玆咅勃私山 丨而忒多媒體鏡即是藉由上述裝 / -影像或音訊t峨的顧,提供—互動式數位料===並回饋 【實施方式】 以下配合圖示對本創作做進 第-圖係代表本創作之多媒明後當更能明瞭。 面單元101,設置於該多媒體鏡】最^圖=多媒體鏡1包括-鏡 別設置一影像擷取裝置⑽、-音訊擷$ 鏡面單元m後分 置辦、-影像輸出裝置1〇5、 ⑽、-影音辨識處理裝 取裝謂、物崎請、 6 M328569 裝置106係侧魏連結於該影音職處縣置_。 其中’因為該鏡面單元102前後兩面的 單元102前後兩面的光度環境不同 =貝_ ’當該鏡面 越多,光線強度微弱之-面,雖會有部份面:光反射會 但其光度微弱,使該光線強度較強之一面將盔:看;::::几1〇2 ’ 因此使該多媒體鏡可提供一鏡面效果 === 或半反射玻璃、隔熱紙、或聚乙稀製品。早凡碰可為全反射 /而該,像擷取裝置搬可為CCD或攝影機等褒 一影像訊號,而該音訊擷取裝置1〇3可為麥 ς ° 訊訊號’而該多媒體鏡i即是藉由該影像她=以擷取一曰 f:f 103; 104 ’而該影音辨識處理裝置1G4巾 、日辨識處理衣^ =庫可配合提供-資料進行該影像Ϊ3該^ 識是採由 =於,音辨識處理裝 3:=5,聲學_進行分析,完成該音訊訊號辨識;若 輪_======將藉由該影像 如上ί同一圖所不’係代表本創作之多媒體鏡功能示意圖; 1二鏡1可藉由設置該鏡面單s1G1於其最前方而提供 置102 立^夕媒體功能3 ;而該多媒體功3係藉由該影像擷取裝 輸至該景1分別揭取—影像或—音訊訊號,並傳 p曰辨哉處理U 104進行辨識處理,當辨識處理完成後,該 7 M328569 3 =提供如娛樂功能4、個人化功能5、盘 現是藉由該影音辨識處理衫、^將^ =:重 ====音訊輪咖= ===能;提供包括多媒雜影 .取個人娜,可提供讀 當選取執行該資訊化功能6時,A :或:子,二牛11,或 輸出ti 乡舰鏡1岐齡斬齡輸人訊號並 翰出對就果’進而提供數位影音多媒體互動式服務。 鏡之圖並;二三圖,表本創作多媒體 疋預先於該影音辨識處理裝置1G4之資料庫中儲存-隱藏式馬可夫模型(HiddenMark〇vMQdel,_進 2〇ϊ辨‘ii貞'巧娜18、性別辨識19、與語言學解碼 果Μ °日/瓜王進仃为析,而完成該音訊訊號的辨識而可得到一辨識結 其中’,隱藏式馬可夫模型(Hidden Mark〇v Mc)del,麵)是_ 辨:ΐϊϊ事件狀態所提出的統計結構,該觀可有效提升了語音 =的正轉及建立μ好之不歡語者(Speaker—論卿耐)的 H’ f此該聲學模型被大量應用於在語音辨識與回應的技術領域 —,I亥^又偵測17是當該多媒體鏡偵測收錄到一語音訊號16時, 猎由:十^吾音的過零率(zer〇 cr〇ssing,娜)和能量值(驗勒) 辨識該語音訊號16是否為一正確交談或命令,·該參數擷取18,是用 f處理辨識語音這種高度差異的訊號時,去找到具有鑑別度的特徵來 數,利用梅爾頻率倒頻譜參數(Mel Freq職cy Cepstral c〇efficients : )及茶數平滑技術(mean subtraction,variance normalization,and 8 M328569 ίο股票 11電子郵件 12天氣 13時間 14新聞 15網際網路 16語音輸入 17聲音段偵測 18參數擷取 19性別辨識 20語言學解碼 21辨識結果 22二維影像序列 23唇形變化圖 24顏色分割 25動態分割 26特徵點擷取 27三維座標 28聲波 29同步處理 30影音回饋M328569 VIII. New Description: [New Technology Field] This creation is about a multimedia device, especially for multimedia mirrors that provide interactive digital audio and video multimedia services. [Prior Art] With the rapid development of technology, the development of digital homes, the creation of smart life and market trends, and in the application of digital home appliances, digital audio and video devices gradually become more powerful: , has patents related to mirror and display functions, i lilt aT〇7563 /Αν"1#ϋ crystal display' and simultaneously used for display purpose mirror module =, self-display μ first transmission, and can indeed come from the screen曰 outside the area: = Knowing the polarizer 'Almost allows light to go through the lens surface, so that the program can cry and force the 'mirror' in the state of use - display, and in the heart The next is only a face mirror; but the polarizer's _ ^ uses the same mass'. The mirror module only provides the surface and the popularity can not be provided - the silk ugly video multi-ship service: into the buckle material to output audio and video, [ New content] II As mentioned above, this creation provides a kind of multimedia service. The multimedia mirror includes a mirror-single interactive digital farm, a video recognition processing and loading device, and an audio recording i-connection relationship. Handling the installed audio wheel, The translucent transposition and the noisy 彡 彡 梅 梅 , , , , , , , , , , , , , , 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜 镜'The light intensity is weak - the surface, although some light will penetrate the mirror, the luminosity of the door is weak, so that the intensity of the county line is stronger - the surface will look at 1f70', but the multimedia mirror created by this The principle is executed, the image, and this will be higher than the light position outside the multimedia mirror, (4) reach == the strength of the plant source and the body mirror can be ah with Wei (four) can show Wei. Therefore, the image capturing device, the audio capturing device, and the output device are electrically connected to the audiovisual recognition processing device, and the audio device can be electrically connected to the tree (awge C() Upled Devi; e, 'Green 撷 置 ' is used to capture - video signals, and the audio capture device ^ machine and other device for picking up, will take (four) take ^ to compare and identify the audio, or The database is also a signal that can be provided by the disciplinary library or processed by the audio processing device. The signal will be sent back to the image through the shadow i. The multimedia mirror is provided by the above-mentioned device / - image or audio t峨, providing - interactive digital material === and feedback [implementation] After the first picture is representative of the multi-media of the creation, it is more clear. The face unit 101 is set in the multimedia mirror. The most picture = the multimedia mirror 1 includes - the mirror is set to an image capturing device (10), - audio 撷$ Mirror unit m after the office, - image output device 1 〇 5, (10), - video recognition The loading and unloading, Ishizaki, and the 6 M328569 device 106 are connected to the audio and video office. _ where 'because the specular unit 102 has different photometric environments on the front and rear sides of the unit 102 on both sides of the mirror unit 102 = _ ' The more the mirror surface, the weaker the light intensity, the surface, although there will be some surface: the light reflection will be weak but the luminosity is weak, so that the light intensity is stronger on one side of the helmet: see;::::a few 1〇2 ' The multimedia mirror can provide a mirror effect === or semi-reflective glass, heat-insulating paper, or polyethylene. The early touch can be totally reflective / and the pick-up device can be a CCD or a camera. The video signal, and the audio capture device 1〇3 can be a microphone signal ', and the multimedia mirror i is obtained by the image======================================================================== The device 1G4 towel, the Japanese identification processing clothing ^ = library can cooperate with the supply-data to perform the image Ϊ3, the knowledge is obtained by =, the sound recognition processing device 3:=5, the acoustic _ analysis, complete the audio signal identification; The wheel _====== will represent the multimedia of this creation by the image as above Functional diagram; 1 2 mirror 1 can provide a 102-litre media function 3 by setting the mirror single s1G1 at the forefront; and the multimedia function 3 is separately extracted by the image capture to the scene 1 Take the image or audio signal and pass the U 104 to identify the processing. When the identification process is completed, the 7 M328569 3 = provide entertainment function 4, personalization function 5, and the disc is played by the video. Identification processing shirt, ^ will ^ =: heavy ==== audio wheel = === can; provide multimedia artifacts. Take the individual, can provide reading when selecting to perform the information function 6, A: or : Zi, Er Niu 11, or output ti xiangship mirror 1 year old age loses the signal and Han out of the fruit, and then provides digital audio and video interactive services. Mirror image; two or three maps, the table creation multimedia 疋 pre-stored in the database of the audio-visual recognition processing device 1G4 - hidden Markov model (HiddenMark 〇 vMQdel, _ into 2 〇ϊ ' 'ii贞' Qiao Na 18 , gender identification 19, and linguistic decoding fruit Μ ° day / melon king into the analysis, and complete the identification of the audio signal can get a recognition of the ', hidden Markov model (Hidden Mark 〇 v Mc) del, face) Yes _ Discrimination: The statistical structure proposed by the event state, which can effectively improve the forward rotation of the speech = and establish the H'f of the unsatisfied person (Speaker-On Qing Nai). Applied to the technical field of speech recognition and response - I Hai ^ and detection 17 is when the multimedia mirror detects a voice signal 16 recorded, the hunting zero: the zero crossing rate of the ten sounds (zer〇cr〇 Ssing, na) and energy value (recognition) to identify whether the voice signal 16 is a correct conversation or command, the parameter is 18, is to use f to process the highly different signal of the recognized speech, to find the discriminant The characteristics of the number, using the Mel frequency cepstral parameters ( Mel Freq job cy Cepstral c〇efficients : ) and tea number smoothing technique (mean subtraction, variance normalization, and 8 M328569 ίο stock 11 email 12 weather 13 time 14 news 15 internet 16 voice input 17 sound segment detection 18 parameters Capture 19 gender identification 20 linguistic decoding 21 identification results 22 2D image sequence 23 lip shape change 24 color segmentation 25 dynamic segmentation 26 feature points capture 27 three-dimensional coordinates 28 sound waves 29 synchronization processing 30 video feedback