201025149 I l X ▼▼ w··/-Ti J 丨 九、發明說明： 【發明所屬之技術領域】 本發明是有關於一種辨識方法及其裝置，且特別是有 關於一種圖形之辨識方法及其裝置。 【先前技術】 徒手繪圖是一種自然且方便的訊息傳遞方式。在替代 文字訊息、操作控制、和人機互動等應用上，徒手繪圖有 ❹ 很大的發展潛力。特別是幾何圖形與符號，它們相較於文 字，更具有直覺、通用，且容易學習與記憶的優點。 若將傳統的字元與符號的辨識系統中應用於圖形的 辨識的話，由於傳統的辨識系統大多是利用訊號處理的技 術，以取得角度、斜率、曲率、頻率響應等特徵值的方式， 透過特徵資料庫來進行辨識。有些傳統作法並透過學習引 擎來熟悉使用者的書寫習慣，以減少判斷之誤差。然而， 這樣的系統卻具有複雜度高，成本高昂，運算量龐大，處 _ 理時間長之缺點。如何發展出效率高且成本低之圖形辨識 之方法，實乃業界所致力的方向之一。 【發明内容】 本發明係有關於一種圖形之辨識方法及其裝置，具有 低成本，運算量少，辨識快速之優點。 根據本發明，提出一種圖形之辨識方法，包括下列步 驟。擷取一軌跡。從軌跡取得多個特徵點。對此些特徵點 201025149 進行排序，並據以對此此 小於-合併誤差值之兩個^徵點進行編號。若存在有距離 產生此些特徵點之―特徵^點，則將兩個特徵點合併。 關係表，特徵點連線關、= 點連線關係表與1徵點位置 連線關係，而特徵錄了兩兩特徵點之間是否有 二關 線關係表與多個待擇圖形7個待選擇連 —ML _表進打比對，以辨識出軌 根據本發明，更接4j ^ 包括-輸入單元、2 用以執行圖形辨識之裳置， 庫LH;單:儲::r-處理單元、-圖形資料 徵點，對此此特徵早元用以從軌跡取得多個特 ;了 一行做點進行排序，並據以對此 :說。當存在有距離小於一合併誤差值=行 將㈣小於—合併料值之兩個特徵點合併。2徵，時’ 用以產生此些特徵點$ ’處理單元更 位置關係表。特徵表與-特徵點 錄兩兩特徵點之間的表:根據多個方向來記 館存多個待選擇圖形的;=連=:#料庫單元用以 選择位置關係表。辨— 關係表與多個待 與特徵點位置關係表 選擇位置關係表進行比對，以辨識出軌=== 201025149 Γ^Λ 為讓本發明之上述内容能 實施例，並配合所附圖式，作詳細說明^下了文特舉較佳 【實施方式】 之辨二其::乃本發明-實施例之 接著，執杆:知圖。首先，執行步驟102, _ 圖形 ❹ 行步碌ι :二從執跡取得多個特徵點。秋一後執跡。 點進行端C進行排序，雜‘t執 合併講差估 後，步驟108尹，若存為古 4特徵 接著執行點蓋則將兩個特徵點合併離小於一 C表與1徵點位置關係J此=徵點之1徵點連線 二::?之間是否有連線二==:係表記錄 :多：方向來記錄兩兩特= 立置關係 112 _ 係表與多個：別與多個待選擇圖形的多個待、.技特徵點 對應“:待選擇位置關係表進行比對，::¾ 些筆^述之轨跡例如具有多個筆蚩。上址少 據這:::折二終二 若r:=:是根據、 產生〜連線_:::二併後 7 201025149 1 間的連線關係’而特徵點位置關係表則是記錄了合併後之 兩兩特徵點之間的相對位置關係。茲舉一例詳細說明如 下。 ’ 請參照第2A圖至帛2C圖，帛2A圖緣示本實施例之 軌跡之一例’第2B圖繪示第2A圖之軌跡所對應之多個特 徵點之分佈’第2C圖緣示合併後之多個特徵點之分佈。 假設使用者於一輸入裝置上畫出了如第2a圖所示之軌跡 200，軌跡200具有筆晝a、筆畫b與筆畫c。假設使用者 係依照箭頭所示之方向來畫出筆畫a、b與c。 於步驟1〇2之擷取軌跡200之後，本實施例之方法接 著執行步驟iG4 ’從軌跡取得多個特徵點，例如是特 徵點al、a2、Μ、b2、cl及c2。特徵點及&分別為 筆畫a之起點與終·點，特徵點Μ * b2分別為筆晝匕之起 點與終點，而特徵點Cl及c2則分別為筆晝c之起點與終 - ' 叫识。空付傲點ai、a2、bl、1 cl及c2進行排序。於本例中，此排序的動作係根據碎 點al、a2、b卜b2、cl及e2之橫座標值與縱座標 行排序。 μ " 舉例來說’先根據特徵點al、a2、bl、b2、Μ 之橫座錄’例如是X座標值，來進行排序，以 al、a2、bl、b2 el及e2依序編號。假設座絶$ c2 ι 點位於左上角。若依照X座標值的大小W而大f 話，排序後的結果將為特徵點a2、^、al、M __ 201025149 f Λ TT t % 此時’分別將特徵點a2、cl、al、Wd及Μ 徵點編號0、勹幵 然後，再根據特徵點al、a2、bl、b2、cl及c2之縱 座標值，例如是γ座標值，將實質上位於相同直線上之至 ^兩個此些特徵點進行排序’以重新對此些特徵點依序編 點的同直線上的定義為’若兩個特徵 ❹ ❿ 直線上，疋義則特徵點a2與cl位於同一條 3上而特徵點b2與C2則位於另一條直 條 直線係沿著縱向延伸。 力條1線上。上边之 值的Π若將位於同—條直線之特徵點，依日3 Y庙押 值的大小由小而大排 伙…、Y座標 點ci之前，然而特徵點b2ma2將仍然位於特徵 如此，重新排序後的結果料^成位於特徵點之前。 及 $ 此時，___a2特徵 ^2、Cl、al、bl、b2 別編號為特徵點編號〇=點 於步驟1〇8中，— 2、3、4及5。 ::點’則將兩個特徵點:二距::::併誤差值之兩個 :距離m均小於合併誤離:2 ’以及特徵點b2及c2 為特徵點a】，胺必、值’故將特徵點al斑hi人 點al、將特徵點a2# 與Μ合併 與C2合併為特徵點b2，。而二5併4 a2,，將特徵點b2 此之間的距離均大於合併誤而差H政點al，、C及彼 誤差值，故不再對特徵點al，、 9 201025149 a2’及b2’進行合併。 依照特徵點a2、cl、al、bl、b2及c2分別編號為特 徵點編號0、1、2、3、4及5，特徵點&2’、&1’及52’的編 號將分別為特徵點編號〇、1及2。 接著，於步驟110中，本方法將產生特徵點al’、a2’ 及b2’之特徵點連線關係表與特徵點位置關係表。請參考 第3A〜3D圖，第3A圖顯示特徵點al’、a2’及b2’之連線 關係與相對位置關係，第3B圖顯示八個方向與其所對應 之方向編號，第3C圖顯示第3A圖之特徵點al’、a2’及 b2’之特徵點連線關係表，而第3D圖則顯示了第3A圖之 特徵點al’、a2’及b2’之特徵點位置關係表。 由第2A及2B圖可知，由於特徵點al與a2有連接 關係，所以合併特徵點al與bl後得到之特徵點al’與合 併特徵點a2與cl後得到之特徵點a2’有連接關係。由於 特徵點bl與b2有連接關係，所以合併特徵點al與bl後 得到之特徵點al’與合併特徵點b2與c2後得到之特徵點 b2’有連接關係。由於特徵點cl與c2有連接關係，所以合 併特徵點a2與cl後得到之特徵點a2’與合併特徵點b2與 c2後得到之特徵點b2’有連接關係，如第3A圖所示。 如此，若將有連接關係者標示為1，而無連接關係者 標示為0的話，則可得到第3C圖所示之之特徵點al’、a2’ 及b2’之特徵點連線關係表。 步驟110中之多個方向例如是八個方向。於第3B圖 中，八個方向分別為上方、右上方、右方、右下方、下方、 左下方、左方及左上 於第3D _之第3 ’其分別標示為方向 置關係表中，若之特徵點al，、a2，及 至8。 看，則特徵點編號Γί點編號0之特徵點 特徵_分別位於之二徵.點ar及特徵點m 推動圖之特別填入方向編號。 請參考第Μ 表之其他内容值1 ::方向之判斷方式的示;;產==置關係表時之 ::-種判斷方式係如下所述。特徵點的相對方 之中心點，另-特徵點為Q1至徵點Ρ位於第4圖 =標值之差值小於-縱座標值差值：若:個特徵點之 之相對位置關係為左 *值〇1?2’則兩個特徵點 值與特徵點P的Y座二估特徵點Q1的Y座榡 且特徵m 小於㈣標值差值_， 顿點Q1的X座標值大於特徵點p的 籲 為位於。的右方。同理可知二： 則位於P的左方。 若兩個特徵點之橫座標值之差值小於—橫座標值差 F3，則兩個特徵點之相對位置關係為上方或下方。例 如特徵點Q3的X座標值與特徵點p的χ座標值的差值小 於橫座標值差值DF3 ’且特徵點q3的γ座標值大於特徵 點Ρ的γ座標值，故特徵點q3將被視為位於ρ的上方。 同理可知，特徵點Q4則位於P的下方。 除了上次四種情況之外，基於兩個特徵點之橫座標值 201025149 的大小關係與縱座標值的大小關 置關係為右上方、右下方、左下方、〆特徵點之相對位 點Q5的X座標值與特徵點Μ =左上方。例如特徵 標值差值DF3，且特徵點Q5的 *值的差值大於橫座 座標值的差值大於縱座標值差值標值與特徵點p的γ 座標值與γ座標值分別大於特徵點/且特徵點如的X 標值，故特徵點Ω5將被視為位於ρ 、座標值與Υ座 特徵點Q6、Q7與Q8分別位於' 的右上方。同理可知， 右下方。 左上方、左下方、與 於步驟112中，第3C圖之特徵机± ^ 圖之特徵點位置關係表將分別二，連線關係表與第 待選擇連線關係表與多個待選擇位mu®形的多個 辨識出執跡所對應之圖形。例如，’、仃纟對，以 特徵點連線關係表與第3D圖^找出與第3C圖之 之多個待選擇連線_表之與特f置關絲最接赴 之一，+ & 與夕個待選擇位置關係表 以完成圖:=之顧擇圖形作為軌跡2。。之圖形， 請參照第5圖，其緣示勃杯贫 -^ w . -λ-iA ® 仃第1圖之圖形之辨識方法 括、 之-例。㈣執行_形賴之裝置500包 :::入单元5〇2、-儲存單元MM、-處理單元506、!： 形資科庫單疋5〇8及一辨識單元510。輸入單元5〇2用 =操取軌跡:儲存單元5G4 ^儲存此軌跡之資訊。處現 早το 506用以從上述之軌跡取得多個特徵點，對此些特徵 點進行排序，並據以對此些特徵點進行編號。當存在有矩 12201025149 I l X ▼▼ w··/-Ti J 丨 、, invention description: [Technical field of the invention] The present invention relates to an identification method and apparatus thereof, and in particular to a method for identifying a pattern and Device. [Prior Art] The freehand drawing is a natural and convenient way of transmitting messages. In alternative text messaging, operational controls, and human-computer interaction applications, there is a great potential for hand-drawn drawings. In particular, geometric figures and symbols are more intuitive, versatile, and easier to learn and remember than text. If the traditional character and symbol recognition system is applied to the recognition of graphics, most of the traditional identification systems use signal processing techniques to obtain feature values such as angle, slope, curvature, and frequency response. The database is used for identification. Some traditional practices use the learning engine to familiarize the user's writing habits to reduce the error of judgment. However, such a system has the disadvantages of high complexity, high cost, large amount of computation, and long processing time. How to develop a high-efficiency and low-cost method of graphic identification is one of the directions that the industry is working on. SUMMARY OF THE INVENTION The present invention relates to a method for identifying a graphic and an apparatus thereof, which have the advantages of low cost, low computational complexity, and fast identification. According to the present invention, a method of identifying a figure is provided, comprising the following steps. Take a track. Get multiple feature points from the trajectory. The feature points 201025149 are sorted, and the two points that are smaller than the - merge error value are numbered accordingly. If there is a feature that produces a feature point for the feature points, the two feature points are merged. Relationship table, feature point connection off, = point connection relationship table and 1 point position connection relationship, and feature records whether there are two off-line relationship tables and multiple candidate patterns between the two feature points Selecting the connection - ML _ table to compare to identify the derailment according to the invention, further 4j ^ includes - input unit, 2 for performing graphic recognition, library LH; single: storage:: r-processing unit, - Graphic data points, for which this feature is used to obtain multiple specials from the trajectory; a row is sorted by points, and according to this: say. When there are two feature points whose distance is less than a combined error value = line (4) is less than - the combined material value is merged. 2 sign, time ' is used to generate such feature points $ ' processing unit more position relationship table. The feature table and the feature point record the table between the two feature points: the plurality of directions are recorded according to the plurality of directions; the ===# repository unit is used to select the position relationship table. </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT [Embodiment] The second embodiment of the present invention is: followed by the present invention - the embodiment, the stick: know. First, step 102 is executed, _ graphics 步 step ι: two obtains a plurality of feature points from the trace. After the fall, he was on the road. The point is sorted by the end C, and after the miscellaneous 't is combined with the difference estimation, the step 108 is Yin. If the save is the ancient 4 feature and then the point cover is executed, the two feature points are merged away from the positional relationship of less than one C and one sign. This = sign of the point 1 connection line 2: :? Is there a connection between two ==: table record: more: direction to record two two special = stand relationship 112 _ series and multiple: not with A plurality of to-be-received and technical feature points of the plurality of to-be-selected graphics correspond to: "The positional relationship table to be selected is compared, and:: 3⁄4 the strokes of the pens are, for example, having a plurality of pens. The upper address is less according to this: :Fold two final two if r:=: is based on, generate ~ connection _::: two and then 7 201025149 1 between the connection relationship 'and the feature point position relationship table is recorded after the merged two two feature points The relative positional relationship between the two is described in detail below. 'Please refer to the 2A to 2C diagrams, the 帛2A diagram shows an example of the trajectory of the embodiment, and the 2B diagram shows the trajectory of the 2A diagram. The distribution of the plurality of feature points 'the 2C picture shows the distribution of the plurality of feature points after the combination. Suppose the user draws an image on the input device as shown in FIG. 2a. The trajectory 200 has a stroke 昼 a, a stroke b and a stroke c. It is assumed that the user draws the strokes a, b and c according to the direction indicated by the arrow. After the trajectory 200 is taken in step 1 〇 2, The method of the embodiment then performs step iG4' to obtain a plurality of feature points from the trajectory, such as feature points a1, a2, Μ, b2, cl, and c2. The feature points and & are the starting point and the ending point of the stroke a, respectively. Point Μ * b2 is the starting point and ending point of the pen ,, and the feature points C1 and c2 are the starting point and ending ' of the pen 昼 c respectively. The empty paying points ai, a2, bl, 1 cl and c2 Sorting. In this example, the sorting action is sorted according to the abscissa value and the ordinate row of the broken points al, a2, b, b2, cl, and e2. μ " For example, 'based on the feature points a, a2 , bl, b2, Μ horizontal coordinate record 'for example, the X coordinate value, to sort, with al, a2, bl, b2 el and e2 sequentially numbered. Assume that the $ c2 ι point is located in the upper left corner. The size of the coordinate value is W and the size of f is large. The sorted result will be the feature points a2, ^, al, M __ 201025149 f Λ TT t % Point a2, cl, al, Wd, and Μ sign number 0, 勹幵 then, according to the ordinate values of the feature points a1, a2, bl, b2, cl, and c2, for example, the γ coordinate value, will be substantially the same On the straight line to ^ two of these feature points are sorted 'to re-define the feature points on the same line as the definition of 'if two features ❹ 直线 on the line, 疋 meaning the feature points a2 and cl are located On the same strip 3, the feature points b2 and C2 are located in the other straight line and extend in the longitudinal direction. Force bar 1 line. If the value of the upper edge is located at the characteristic point of the same line, the size of the 3 Y temple is based on the size of the small and large squad..., the Y coordinate point ci, but the feature point b2ma2 will still be located in the feature, re The sorted result is placed before the feature point. And $ At this time, the ___a2 feature ^2, Cl, al, bl, and b2 are numbered as feature point numbers 〇=points in steps 1〇8, — 2, 3, 4, and 5. ::Point' then two feature points: two distance:::: and two error values: distance m is smaller than the combined error: 2 ' and feature points b2 and c2 are feature points a], amine must, value 'Therefore, the feature point a spot is the person point a, the feature point a2# is merged with the Μ and C2 is merged into the feature point b2. And two 5 and 4 a2, the distance between the feature points b2 is greater than the combined error and the difference H, a, C and the error value, so no longer the feature points a,, 9 201025149 a2 ' and b2 'Merge. According to feature points a2, cl, al, bl, b2, and c2, they are numbered as feature point numbers 0, 1, 2, 3, 4, and 5, respectively. The number of feature points & 2', & 1' and 52' will be respectively The feature points are numbered 〇, 1 and 2. Next, in step 110, the method generates a feature point connection relationship table and a feature point position relationship table of the feature points a', a2', and b2'. Please refer to FIG. 3A to FIG. 3D. FIG. 3A shows the connection relationship and relative position relationship of the feature points a', a2' and b2', and FIG. 3B shows the direction directions of the eight directions and the corresponding direction, and FIG. 3C shows the first The feature point connection relationship table of the feature points a', a2', and b2' of the 3A figure, and the 3D figure shows the relationship table of the feature points of the feature points a', a2', and b2' of the 3A figure. As can be seen from Figs. 2A and 2B, since the feature points a1 and a2 are connected, the feature points a' obtained by combining the feature points a1 and bl are connected to the feature points a2' obtained by combining the feature points a2 and cl. Since the feature points bl have a connection relationship with b2, the feature points a' obtained by combining the feature points a1 and bl are connected to the feature points b2' obtained by combining the feature points b2 and c2. Since the feature points cl and c2 are connected, the feature points a2' obtained by combining the feature points a2 and cl are connected to the feature points b2' obtained by combining the feature points b2 and c2, as shown in Fig. 3A. Thus, if the connection relationship is indicated as 1 and the connectionless relationship is marked as 0, the feature point connection relationship table of the feature points a', a2', and b2' shown in Fig. 3C can be obtained. The plurality of directions in step 110 are, for example, eight directions. In Fig. 3B, the eight directions are upper, upper right, right, lower right, lower, lower left, left, and upper left at the 3rd _th 3', which are respectively indicated as the direction relationship table, if The feature points are al, a2, and 8. See, the feature point number Γ ί point number 0 of the feature points feature _ respectively located in the second sign. Point ar and feature point m push the chart into the special fill direction number. Please refer to the other contents of Table 值1::::::::==================================================== The center point of the opposite side of the feature point, the other - the feature point is Q1 to the sign point Ρ is located in the fourth picture = the difference between the values of the scale is less than - the value of the ordinate value: if: the relative positional relationship of the feature points is left * The value 〇1?2' is the two feature point values and the Y-seat of the feature point P. The Y-station of the feature point Q1 is evaluated and the feature m is smaller than the (four)-value difference _. The X coordinate value of the punctual point Q1 is larger than the feature point p. The appeal is located. On the right. The same reason can be seen two: it is located to the left of P. If the difference between the abscissa values of the two feature points is less than the abscissa value difference F3, the relative positional relationship between the two feature points is above or below. For example, the difference between the X coordinate value of the feature point Q3 and the χ coordinate value of the feature point p is smaller than the abscissa value difference DF3 ' and the γ coordinate value of the feature point q3 is larger than the γ coordinate value of the feature point ,, so the feature point q3 will be It is considered to be above ρ. Similarly, the feature point Q4 is located below P. Except for the last four cases, the relationship between the size of the horizontal coordinate value 201025149 based on the two feature points and the size of the ordinate value is the upper right, lower right, lower left, and relative position Q5 of the feature point. The X coordinate value and the feature point Μ = upper left. For example, the characteristic value difference DF3, and the difference between the * values of the feature points Q5 is greater than the difference between the horizontal coordinate values and the ordinate values of the ordinate values and the γ coordinate values of the feature points p and the γ coordinate values are greater than the feature points, respectively. / and the characteristic point is such as the X value, so the feature point Ω5 will be regarded as located at ρ, the coordinate value and the squat feature points Q6, Q7 and Q8 are located at the upper right of '. Similarly, the bottom right. In the upper left and lower left, and in step 112, the feature point position relationship table of the feature machine of the 3C figure will be respectively two, the connection relationship table and the to-be-selected connection relationship table and the plurality of to-be-selected positions mu A plurality of shapes of the shape recognize the pattern corresponding to the trace. For example, ', 仃纟 ,, with the feature point connection table and the 3D map ^ find out with the 3C map of the plurality of to-be-selected connections _ table and special f to the silk, one of the best, + & and the selected positional relationship table to complete the map: = the choice of the graph as the track 2. . For the figure, please refer to Fig. 5, which shows the identification method of the figure of Fig. 1 of the Bo's cup -^ w . -λ-iA ® 仃 Fig. 1 . (4) Executing _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The input unit 5〇2 uses = to manipulate the track: the storage unit 5G4 ^ stores the information of the track. The το 506 is used to obtain a plurality of feature points from the above trajectory, and the feature points are sorted, and the feature points are numbered accordingly. When there is a moment 12
L 201025149 • I 1 VY 7-¾. 離小於合併誤差值之兩個特徵點時，將距離小於合併誤差 值之該兩個特徵點合併。處理單元506更用以產生此些特 徵點之特徵點連線關係表與特徵點位置關係表。 圖形資料庫單元508用以儲存多個待選擇圖形的多 個待選擇連線關係表與多個待選擇位置關係表。而辨識單 元510則是用以將特徵點連線關係表與特徵點位置關係表 與多個待選擇連線關係表與多個待選擇位置關係表進行 比對，以辨識出上述之軌跡所對應之圖形。 • 用以執行圖形辨識之裝置500例如可應用於數位教 學系統、互動式展示系統、遠端視訊系統、數位電子式互 動遊戲系統中。軌跡之來源可為滑鼠、觸控板、或是影像 特徵體感應裝置。影像特徵體感應裝置例如是可對特定物 體之特定部位進行追蹤之感應器，例如是可對運動中之使 用者肢體進行追蹤之感應器。舉例來說，輸入單元502將 根據滑鼠的移動路徑，來擷取軌跡。或者，輸入單元502 將根據運動中之使用者肢體的移動路徑，來擷取執跡。 ® 本發明之圖形之辨識方法及其裝置透過分析特徵點 之間的連線關係與相對位置關係，來進行圖形的辨識或符 號的辨識，例如是幾何圖形的辨識。由於本發明係以特徵 點之間的連線關係與相對位置關係作為圖形或符號的特 徵資訊，此種辨識方式將不會受到筆晝之順序或筆晝數之 多寡的限制。亦即，同樣的一個圖形，使用者不管用何種 筆晝順序，或是以多少的筆畫數完成，只要特徵點具有類 似的連線關係與相對位置關係，即可得到相同的辨識結 13 201025149 i w3u^r/\ . 果。 此外，本發明更具有高容錯力，與無須使用學習引擎 即可完成辨識之優點。而且，由於僅需產生特徵點之間的 連線關係與相對位置關係，並進行比對，本發明不需如傳 統作法般，必需用到與角度、斜率、曲率、頻率響應相關 之複雜的運算，故可達到運算簡單，處理速度快，低成本 與低系統複雜度之優點。 綜上所述，雖然本發明已以較佳實施例揭露如上，然 其並非用以限定本發明。本發明所屬技術領域中具有通常 © 知識者，在不脫離本發明之精神和範圍内，當可作各種之 更動與潤飾。因此，本發明之保護範圍當視後附之申請專 利範圍所界定者為準。 【圖式簡單說明】 第1圖繪示本發明一實施例之一種圖形之辨識方法 的流程圖。 第2A圖繪示本實施例之軌跡之一例。 ⑩ 第2B圖繪示第2A圖之軌跡所對應之多個特徵點之 分佈。 第2C圖繪示合併後之多個特徵點之分佈。 第3A圖顯示特徵點al’、a2’及b2’之連線關係與相對 位置關係。 第3B圖顯示八個方向與其所對應之方向編號。 第3C圖顯示第3A圖之特徵點al’、a2’及b2’之特徵 14 201025149 I I 1. VT ^V/-»-TX ΓΛ. 點連線關係表。 第3D圖顯示了第3A圖之特徵點al’、a2’及b2’之特 徵點位置關係表。 第4圖繪示當產生特徵點位置關係表時之一種方向 之判斷方式的示意圖。 第5圖繪示執行第1圖之圖形之辨識方法之裝置之方 塊圖之一例。 ❿ 【主要元件符號說明】 102〜112 :流程步驟 200 :軌跡 500:用以執行圖形辨識之裝置 502 :輸入單元 504 :儲存單元 506 :處理單元 508 :圖形資料庫單元 • 510 :辨識單元 15L 201025149 • I 1 VY 7-3⁄4. When the two feature points smaller than the combined error value are separated, the two feature points whose distance is smaller than the combined error value are combined. The processing unit 506 is further configured to generate a feature point connection relationship table and a feature point position relationship table of the feature points. The graphic database unit 508 is configured to store a plurality of to-be-selected connection relationship tables and a plurality of to-be-selected position relationship tables of the plurality of to-be-selected graphics. The identification unit 510 is configured to compare the feature point connection relationship table and the feature point position relationship table with the plurality of to-be-selected connection relationship tables and the plurality of to-be-selected position relationship tables to identify the corresponding track The graphics. • The apparatus 500 for performing pattern recognition can be applied, for example, to a digital teaching system, an interactive display system, a far-end video system, and a digital electronic interactive game system. The source of the track can be a mouse, a trackpad, or an image feature sensing device. The image feature sensing device is, for example, a sensor that can track a specific portion of a particular object, such as a sensor that can track the limbs of a user in motion. For example, the input unit 502 will retrieve the trajectory according to the movement path of the mouse. Alternatively, the input unit 502 will retrieve the trajectory based on the movement path of the user's limb during exercise. The identification method and device of the graphic of the present invention analyzes the relationship between the feature points and the relative positional relationship to identify the figure or identify the symbol, for example, the identification of the geometric figure. Since the present invention uses the connection relationship between the feature points and the relative positional relationship as the feature information of the figure or the symbol, the identification mode will not be limited by the order of the pen or the number of pens. That is to say, the same figure, the user does not use any pen order, or how many strokes are completed, as long as the feature points have similar connection relationship and relative position relationship, the same identification knot can be obtained 13 201025149 i w3u^r/\ . In addition, the present invention is more highly fault tolerant and can be distinguished without the need to use a learning engine. Moreover, since only the connection relationship and the relative positional relationship between the feature points need to be generated and compared, the present invention does not need to use complicated operations related to the angle, the slope, the curvature, and the frequency response as in the conventional method. Therefore, the advantages of simple operation, fast processing speed, low cost and low system complexity can be achieved. In the above, the present invention has been disclosed in the above preferred embodiments, but it is not intended to limit the present invention. It is common in the art to which the invention pertains, and various modifications and refinements can be made without departing from the spirit and scope of the invention. Therefore, the scope of the invention is defined by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flow chart showing a method for recognizing a pattern according to an embodiment of the present invention. Fig. 2A is a diagram showing an example of the trajectory of the embodiment. 10 Figure 2B shows the distribution of multiple feature points corresponding to the trajectory of Figure 2A. Figure 2C shows the distribution of multiple feature points after the combination. Fig. 3A shows the connection relationship and relative positional relationship of the feature points a', a2', and b2'. Figure 3B shows the direction of the eight directions and their corresponding directions. Fig. 3C shows the features of the feature points a', a2' and b2' of Fig. 3A. 14 201025149 I I 1. VT ^V/-»-TX ΓΛ. Point connection table. Fig. 3D is a table showing the positional relationship of the feature points of the feature points a', a2', and b2' of Fig. 3A. Fig. 4 is a view showing a manner of judging a direction when a feature point position relation table is generated. Fig. 5 is a diagram showing an example of a block diagram of an apparatus for performing the identification method of the pattern of Fig. 1. ❿ [Main component symbol description] 102~112: Flow step 200: Track 500: Device for performing pattern recognition 502: Input unit 504: Storage unit 506: Processing unit 508: Graphic library unit • 510: Identification unit 15