201128463 六、發明說明: 【發明所屬之技術領域】 本發明係有關一種手部軌跡辨識系統及其方法,尤 其是具有手部視訊介面並使用鏈碼為執跡特徵再利用混 淆集合進行修正。 【先前技術】 目前應用在手寫技術方面上,其習知技術為都需要 有另外的輸入設備,如手寫板、觸控面板,這類的設備 車乂為叩貝,使知相關產品的成本提高;此外,過去此類 技術大多需要—隻特㈣筆做為輸人卫具,而這類的筆 大多為輕細短小,容易遺失,也造成操作上有許多不便。 現今許麵控技術已經可㈣手代筆,然而這些觸 =面,健存在著厚度厚、綱面易雨、透紐較差 問題。雖然觸控面板已經逐漸普及於多 型手機、自動樞員機以及部分觸控 、·一f電腦豕電化的目標裡’觸控面板反而顯得不實 類大二觸控?板成本較高,無法直接普及在如電視 反而顯得笨:市f二來,大型面板應用的觸控介面 操作: 為使用者需要往返座位與面板間進行 類似要—種手部執跡辨識系統及其方法,利用 點。觸控式介面的遙控操作,以解決上述制技術的缺 201128463 tBACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a hand trajectory recognition system and method thereof, and more particularly to having a hand video interface and using a chain code as a trace feature to re-use a confusing set for correction. [Prior Art] Currently applied to handwriting technology, the conventional technology requires additional input devices, such as a tablet, a touch panel, and the like is a mussel, which increases the cost of related products. In addition, in the past, most of these technologies were needed—only special (four) pens were used as input aids, and most of these pens were light and short, easy to lose, and caused a lot of inconvenience in operation. Nowadays, the face control technology is already available. (4) Handwriting, however, these touches face, there are problems of thick thickness, easy rain, and poor penetration. Although the touch panel has gradually become popular in multi-type mobile phones, automatic cross-border machines, and some touch-sensitive, one-five computers, the target of the touch panel is not realistic, so the type of sophomore touch is high. Direct popularization, such as TV, is stupid: the city's second two, the touch interface operation of large panel applications: for the user needs to go back and forth between the seat and the panel to do similar - hand hand identification system and its method, the use of points. The remote operation of the touch interface to solve the lack of the above-mentioned system technology 201128463 t
【發明内容】 並方主要目的在提供—種手部執跡辨識系統及 福^ 訊攝影機模組、移動物件偵測模組、手指 =、、、,、且、舰擷取额及触觸顯, 逢盥象,軸物件伽《終合背景重 物體色彩分析技術找出動作中的手部 模組根據分離出的手部景彡傻 m 端細Μ、 手指尖端,並依據手指尖 仏己錄執跡,特徵榻取模、组使用鏈碼做為軌跡 觸触徑向基底函數類神_路對鏈碼 特徵進仃为類’再經混淆集合進行修 跡所表示的觀敎字。 曰罵軌 因此本發明係利用視頻訊號檢測以實現手寫輸入, 可做為電腦動作之依據,進行電腦控制或是文字輸入,進 而改善‘人機溝通的便利择,可應用於電腦家電化、即時 ί腦教學、辦公室的會議、資料輸入、文章撰寫或繪圖所 给的人機介Φ ’尤其是對於語言障礙的病患,在不需透過 任何外在的輸人設備下’即可直翻料指書寫的方式進 行對談與溝通。 【實施方式】 以下配合圖式及元件符號對本發明之實施方式做更 詳細的說明,俾使«該項技藝者在研讀本書後能據 以實施。 參閱第一圖’本發明手部執跡辨識系統的示意圖。如 第一圖所示,本發明的手部轨跡辨識系統包括視訊攝影機 模組10、移動物件偵測模組20、手指偵測模組30、特徵 201128463 擷取模組40及軌跡辨識模組50,用以提供手部視訊介面 以實現軌跡辨識功能。 視5fl攝景>機模組1 〇拍攝畫面以產生複數個且連續的 動態影像資料T1,而視訊攝影機模組1〇可包括電荷耦合 元件(CCD)或互補式金氧半(CM〇s)攝影機,且具有複數個 像素。每一動態影像資料T1包括複數個像素值,而每個 像素具有相對應的像素值。 栘動物件偵測模組20接收動態影像資料T1以建立背 ,景f像’並啊制出祕影像資料T1巾的飾物件T2。 隨著影像的移動及時間的改變,像素值也會同步變動,因 此先對動_影像資料T1的像素值進行統相建立直方圖 p St〇gram) ’再利用核函數密度估測法(Kernel Density 值職)針對每個像素峨生逼近的相對應像素 ^機率’最後依據下一動態影像資料Τ1而由每個像 ㈣該像素是屬於不 移動物件T2,且判斷美進a八说μ 飞移動的[Summary of the Invention] The main purpose of the joint is to provide a hand-existing identification system and a camera module, a moving object detection module, a finger=,,,,,, and a ship's take-up and touch display. , Everything, axis object gamma "final background background heavy object color analysis technology to find out the hand module in action according to the separated hand scene 彡 silly m end fine, finger tip, and according to the tip of the finger Execution, the characteristic couch modulo, the group uses the chain code as the trajectory touch radial base function class _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Therefore, the present invention utilizes video signal detection to implement handwriting input, and can be used as a basis for computer actions, performing computer control or text input, thereby improving the convenience of human-machine communication, and can be applied to computer home appliances, instant脑 Brain teaching, office meetings, data entry, article writing or drawing gives the human-machine interface Φ 'especially for patients with language disorders, without any external input device' Refers to the way of writing to talk and communicate. [Embodiment] Hereinafter, embodiments of the present invention will be described in more detail with reference to the drawings and the component symbols, so that the skilled person can implement the book after studying the book. Referring to the first figure, a schematic diagram of the hand trace recognition system of the present invention. As shown in the first figure, the hand track recognition system of the present invention comprises a video camera module 10, a moving object detecting module 20, a finger detecting module 30, a feature 201128463 capturing module 40 and a track recognition module. 50, used to provide a hand video interface for track recognition. Depending on the 5fl camera> machine module 1 〇 capture picture to generate a plurality of continuous motion picture data T1, and the video camera module 1 〇 may include a charge coupled device (CCD) or a complementary MOS half (CM〇s) a camera with a plurality of pixels. Each motion picture data T1 includes a plurality of pixel values, and each pixel has a corresponding pixel value. The animal detecting module 20 receives the moving image data T1 to create a back image, and creates a decorative object T2 of the secret image data T1. As the image moves and the time changes, the pixel values will also change synchronously. Therefore, the pixel value of the moving image data T1 is firstly integrated to establish a histogram p St〇gram) 'Reuse kernel function density estimation method (Kernel) Density (for duty) is the corresponding pixel for each pixel. The probability is based on the next moving image data Τ1 and each image (4) is the non-moving object T2, and the judging a Mobile
機率小的屬於所需的== 機率大的屬於背景而分佈 手才曰偵測拉組30接收移動物件Τ2,先找出指尖,亦 St二ί利用色彩分析技術過濾移動物件Τ2中非 手部32的影像_,^ 6第二g戶斤;-、再、、二 ,同時刀離出 =32相重疊的交點33,%,35為中心,個:^矩 =為指⑽,輪_糊她== 201128463 τ3,触代轉轉錄觸指尖執跡 Τ4黎金W里”書寫特徵榻取,並產生軌跡特徵 ^签^旦抽取係依據書寫特點,比如筆畫有轉彎、提筆咬The probability is small, the required == probability is large, the background is distributed, the hand is detected, the pull group 30 receives the moving object Τ2, first finds the fingertip, and also uses the color analysis technology to filter the moving object Τ2 The image of the part 32 _, ^ 6 second g household; -, then, two, at the same time the knife leaves = 32 overlapping intersection 33,%, 35 as the center, a: ^ moment = refers to (10), round _ Paste her == 201128463 τ3, touch the transcript to touch the fingertips Τ 4 Li Jin W in the "writing feature couch, and generate trajectory features ^ sign ^ Dan extraction based on writing characteristics, such as strokes have a turn, pen bite
Iff _尖的_速度會減慢,而由指尖軌跡Τ3中 曰曰’、移絲度的轉以取岭頓點 f尖軌射3巾的私«、離騎晝與書寫筆Γ同刀 離開筆晝,僅保留書_以供判斷。 ^ 晝依照其書寫方向進行 fh細Code)編碼’尺為正整數,並利用直方圖對奎寫 ^的軌跡移動方向的機率進行統計,再藉核函數密度估 用仿以,測機率分佈。由於鏈碼的循環編碼特性,造成利 用^^估測法所估測出的函數範圍會落於小於〇或超過 =,_方向為_區域·,相對應之機率必$加 =向〇的區域’此稱為循環式鏈碼機率分佈估測,而估 測出的機率函數係當作觸㈣軌跡特徵T4。 轨跡職模級50接收轨跡特徵T4,進行轨跡辨識, 二生書寫符號Τ5 ’係先依據徑向基底函數類神經網路 跡 Function Neural Networks,RBFNN)對軌 η τΑ π分類,以產生初步分賴果。&於軌跡特 ^中不同的文字或符號會有相似的鏈碼分佈,例如〇 與0 ’ 2與Z ’因此再混賴攸⑽using㈣以修 =刀步分類結果’並產生所需的書寫符號T5,可包括手寫 符號或手寫文字,可做為電麟作之轉,進而實現進行 電腦控制或是文字輸入。 參閱第三圖,本發明手部軌跡辨識方法的示意圖。如 .201128463 S1° ^ 產生連續的複數個動態影像::攝影機模組以拍攝畫面, 包括=像素值,而每個像素值係對應於-像;貝科 等動態影像資料,並!動物件伽|J模組接收該 率大的像辛t背旦=近似的分佈機率,並以分佈機 轉狀^,而分佈機狗、 體’,以_出所需的移動物件,並進入步驟咖 -手指偵職組接收該轉 ] Μ 非手部的移動物體進而二:=術過 部,再以包圍矩形將手部影像的輪廊框起來 包圍矩形相重疊的交點為中心 :σ ” 而具最小夾角的.交點“尖 的連續位置以職齡執跡。 置雜“ 接著在步驟S40中,利用一特徵 執跡,係先藉指尖移動逮度的估算停頓收^尖 停頓點以分難錄跡T3巾的進、用 寫筆畫,將書寫整查昭甘+办—離開筆晝與書 (^Γη Code)^:N;;^^^ 的軌跡移動方向的機率進行統計,再寫筆畫 以估^佈,而估測出的機率錄為軌跡=估測法 最後進入步驟S50,利用一舳 特徵’先依據徑向基底函數類神路; 執跡 :r:=,,再利_集合== 、、,。果,並產生所需的書寫符號。 負 201128463 由於上述本發日月方法的視訊攝影機模組、移動物件谓 :寺徵擷取模組及軌跡辨識模組係 20丰^ 攝频做.移祕件偵測模租 特徵娜模組40伽辨識模組 本發明的特點在於,不須受輸入設備的限制,Iff _ sharp _ speed will slow down, and from the fingertip trajectory Τ 3 曰曰 ', the degree of silk transfer to take the ridged point f tip rail shot 3 towel private «, the riding 昼 and the writing pen Γ knife Leave the pen and keep only the book _ for judgment. ^ 昼 According to the direction of writing, fh fine Code) coded the ruler as a positive integer, and use the histogram to calculate the probability of the direction of the trajectory of the Kui write ^, and then use the kernel function density to estimate the probability distribution. Due to the cyclic coding characteristics of the chain code, the range of functions estimated by the ^^ estimation method will fall below 〇 or exceed =, the _ direction is _region·, and the corresponding probability must be $plus = 〇 'This is called the cyclic chain probability distribution estimation, and the estimated probability function is taken as the touch (four) trajectory feature T4. The trajectory level 50 receives the trajectory feature T4 for trajectory identification, and the second writing symbol Τ5' is first classified according to the radial basal function Neural Neural Network (RBFNN) to the orbit η τ Α π to generate Preliminary results. & different text or symbols in the track feature ^ will have a similar chain code distribution, such as 〇 and 0 ' 2 and Z ' and therefore 混 攸 (10) using (four) to repair = knife step classification results 'and produce the required writing symbols T5, which can include handwritten symbols or handwritten words, can be used as a turn of the electric lining, thereby enabling computer control or text input. Referring to the third figure, a schematic diagram of the method for identifying the hand trajectory of the present invention. Such as .201128463 S1° ^ Generates a continuous plurality of motion pictures: The camera module captures the picture, including = pixel values, and each pixel value corresponds to - image; Becco and other motion picture data, and! The animal piece gamma|J module receives the large rate of distribution like 辛t背旦=approximate distribution probability, and distributes the machine to the machine, and distributes the dog, body', to the desired moving object, and enters Step coffee-finger squadron group receives the turn] Μ Non-hand moving object and then two: = surgery part, and then encloses the rectangle of the hand image framed around the intersection of the rectangles: σ ” The intersection with the smallest angle "points of continuous position is honoured by the age of the staff. "In the next step, in step S40, using a feature to trace, the first step is to use the fingertip to move the catch. The estimated pause is to stop the point and stop the point to divide the difficulty of the T3 towel. Gan + do - leave the pen and book (^Γη Code) ^: N;; ^ ^ ^ the probability of the direction of the trajectory of the movement to statistics, and then write strokes to estimate the cloth, and the estimated probability is recorded as the track = estimate The test finally proceeds to step S50, using a 舳 feature 'first based on the radial basis function classpath; obstruction: r:=, then _set==, ,, and, and produce the desired written symbol. Negative 201128463 The video camera module and the moving object of the above-mentioned method of the present day and the moon are: the temple levy module and the trajectory identification module system 20 Feng ^ video frequency to do. Move the secret detection module rent feature Na module 40 The gamma identification module is characterized in that the invention is not limited by the input device.
手指作為輸人介面 I :=的另一特點在於,針對手部辨識技術, f尖可與手部的輪麟段之最小夾角以定位 ^定位:=Γ:ΓΓ來定位指尖的方法, 的機率分布以作為執跡特徵,並利用徑向 路進行分類,將輸出結果透猶立好的 心有集·"來進行修正’以達到更好的辨識率。 摔作戍文㈣手部軌跡辨識系統可肋控制電腦之 =====便利性’可應用於 文童触德I 學、辦公㈣會議、資料輸入、 赠s,概驗言障礙的 不需透過任何外在的輸人設備下,即可直接利用 手才曰曰寫的方式進行對談與溝通。 企圖僅為用以解釋本發明之較佳實施例,並非 止圖據_本㈣做任何形紅之關 谓作細树仅飾料更,皆仍 應包括在本發明意圖保護之範疇。 201128463 【圖式簡單說明】 第-圖為本發明手部執跡辨識系統的示意圖。 第二圖為本發明手指偵測的示意圖。 第二圖為本發明手部軌跡辨識方法的示意圖。 【主要元件符號說明】 10視訊攝影機模組 20移動物件偵測模組 30手指债測模組 31手部 32包圍矩形 33交點 34交點 35交點 40特徵擷取模組 50軌跡辨識模組Another feature of the finger as the input interface I:= is that for the hand recognition technology, the f-tip can be positioned with the smallest angle of the hand of the wheel to locate the position: =Γ:ΓΓ to locate the fingertip, The probability distribution is used as the detour feature, and the radial path is used for classification, and the output result is transparently set to "to correct it to achieve a better recognition rate.摔作戍文(4) Hand trajectory identification system can control the computer =====convenience' can be applied to Wentong Touche I, office (four) meeting, data input, gift s, no need to test obstacles Through any external input device, you can directly use the way of handwriting to communicate and communicate. The present invention is intended to be merely a preferred embodiment for explaining the present invention, and it is not intended to be a simplification of the invention, and it is intended to be included in the scope of the present invention. 201128463 [Simple description of the diagram] The first figure is a schematic diagram of the hand recognition identification system of the present invention. The second figure is a schematic diagram of finger detection according to the present invention. The second figure is a schematic diagram of the method for identifying the hand trajectory of the present invention. [Main component symbol description] 10 video camera module 20 mobile object detection module 30 finger debt measurement module 31 hand 32 enclosing rectangle 33 intersection 34 intersection 35 intersection 40 feature capture module 50 track recognition module
S10利用視賴影機模組以拍攝畫面產生 個動態影像資料 S20利用移動物件偵測模組接收動態影像 數密度估測法以_移動物件 、错核函 S30利用手指_模組接收移動物件找出最小爽角交 點以形成指尖執跡 S40利用特徵摘取模組接收指錄跡進行筆晝抽取與 書寫特徵擷取以產生軌跡特徵 S50利用軌跡辨識模組接收執跡特徵進行執跡辨識以 產生書寫符號 201128463 T1動態影像資料 丁2移動物件 Τ3指尖執跡 Τ4軌跡特徵 Τ5書寫符號The S10 uses the video camera module to generate a dynamic image data by taking a picture. The S20 uses the moving object detection module to receive the dynamic image number density estimation method. The mobile object and the wrong core S30 use the finger_module to receive the moving object. The minimum refresh angle intersection is formed to form the fingertip representation S40, and the feature extraction module is used to receive the fingerprint track for pen extraction and writing feature extraction to generate the trajectory feature S50. The trajectory identification module is used to receive the trajectory feature for performing the recognition identification. Generated writing symbols 201128463 T1 dynamic image data Ding 2 moving objects Τ 3 fingertips Τ 4 track features Τ 5 writing symbols