201205471 六、發明說明： 【發明所屬之技術領域】 β本發明係舆一種生物特徵锊識方法及系統有關，特別 疋與一種人臉特徵辨識方法與系統有關。 【先前技術】 生物辨識(Biometrics)技術在當前社會扮演著越來越 重要的角色。從提款機、門禁系統、筆記型電腦以至於隨 身碟等應用中，都可以見到生物辨識技術的應用。 # 在生物辨識技術的領域中，人臉辨識技術是一個新興 ，研發重點。習知人臉辨識系統所用的演算法多半是基於 /「核型取特徵」的方式來做特徵擷取。所謂「模型取特 係利用數學降維模型(Dimensi〇n Reducti〇n M〇del)以 杈擬影像。習知常見的方法如主成分分析（principie Component Analysis)或是線性識別分析 D1SCriminant Analysis)等，皆是一種全域(Η_⑽模擬方 法。然而這類習知模擬全域影像的方法，由於是一次取全 财彡像的特性，所赠料受朗取得影像正規化不夠良 好以及一些影像局部破壞(〇cdusi〇n)或是旋轉(r〇加丨如）、 位移（shift)、大小（Scale)以及光#(IUuminati〇n)等破壞影 ，因素所影響。*此類因素皆會影響最後比對的特徵向 里，導致後續特徵向量比對得到不正確的辨識結果。 有鑑於此’本發明揭露一種人臉特徵辨識方法盥系 統’ f係藉由二元化圖形(L〇cal Binary pattem)為基礎， 辅以间斯此合化模型（Gaussian Mixture M〇dd)而模擬取得 特徵點分布資訊，藉以解決習知人臉特徵辨識之問題。 201205471 【發明内容】 本發明之一範疇在於提供一種人臉特徵辨識方法。 根據本發明之-旲體實施例；本發明提供一種人臉特 徵辨識方法，其包含有以下步驟：（S1)自一視訊串流中擷 取一視§fl框，（S2)判斷該視訊框中是否有一人臉影像；（S3) 若有該人臉影像’則將該人臉影像正規化(N〇rmalize); (S4)利用一局部二元化圖形(L〇cai Binary Pattern, LBP)來描 述該正規化後人臉影像之一組特徵；以及(S5)利用一高斯 混合模型（Gaussian Mixture Model，GMM)來模擬該組特徵 之空間分布資訊。 、 、 於貫際應用中，本發明之人臉特徵辨識方法所採用的 局部二元化圖形，係為一種進階多重解析區塊-局部二元 化圖形（Advance Multi-res〇luti〇n Block _ Local Binary Pattern，AMB-LBP)’而步驟（S4)進一步包含有以下子步 驟：（S41)將正規化後之該人臉影像，以一中央像素為中心 分割成複數個區塊，且該複數個區塊係以陣列式排列，而 相鄰之該區塊間具有一預定距離；（S42)利用一中央區塊之 平均灰階值與其相鄰區塊之平均灰階值做局部二元化圖形 運算’藉以產生該中央像素之特徵值；以及(S43)計算該人 臉衫像上之母一像素之特徵值，藉以產生該組特徵。 於實際應用中’該區塊間之預定距離係利用該中心像 素(Center Pixel)以及複數個圍繞該中心像素之周圍像素 (Neighbor Pixels)來定義’而本發明方法之步驟(S4)係藉由 擴增该中心像素與該周圍像素之預定距離，以描述該正規 化後人臉影像之該組特徵，並取得正規化後該人臉影像之 巨觀(Macrostructure)紋理資訊。 201205471 本發明之另一範疇在於提供一種人臉特徵辨識系統。 根據本發明之另一具體實施例’本發明所提供一種人 臉特徵辨識糸統，其包含有—影偉^鎖取模組、一判斷模 組、一轉換模組以及一模擬模組。影像彌取模組係用於自 一視sil串流中掘取一視訊框。判斷模組連接於該影像榻取 模組，以判斷該視訊框中是否有一人臉影像，若有該人臉 影像’則將該人臉影像正規化(Normalize)。轉換模組係與 該判斷模組連接，以利用一局部二元化圖形(L〇cal Binary Pattern，LBP)來描述該正規化後人臉影像之一組特徵。模 擬模組係與該轉換模組連接，以利用一高斯混合模型 (Gaussian Mixture Model，GMM)來模擬該組特徵之空間分 布資訊。 於實際應用中，本發明人臉特徵辨識系統所採用之局 部二元化圖形，係為一種進階多重解析區塊_局部二元化 圖形(Advance Multi-resolution Block - Local Binary Pattern， AMB-LBP)，其係將正規化後之該人臉影像以一中央像素 為中心分割成複數個區塊，而該複數個區塊以陣列式排 列，且相鄰之該區塊間具有一預定距離；接著，利用一中 央區塊之平均灰階值與其之相鄰區塊的平均灰階值進行局 部二兀化圖形運算後，以產生該ψ央像素之特徵值；最 後，計算該人臉影像上每一像素之特徵值，以產生該組 徵。 、’. 於貫際應用中，該區塊間之預定距離係利用該中心像 素(Center Pixel)以及複數個圍繞該中心像素之周圍像素 (Neighbor Pixels)來定義，而本發明系統之轉換模組係藉 由擴增該中心像素與該周圍像素之預定距離，來描述正^ 201205471 化後該人臉影像之該組特徵’並取得正規化後該人臉影像 之巨觀(Macrostructure)纹理資訊。 相較於習知技術，本發明之人臉特徵辨識方法與系 統，係利用局部二元化圖形(LBP)來描述該正規化後的人 臉景^像之一組特徵，接著並利用高斯混合模型（GMM)來 模擬該組特徵的空間分布資訊，藉以建構出可以抵抗偏移 (shift)、些微旋轉(Rotation)、或是尺寸(Scale)變化等影像 破壞因素的特徵。此外，本發明之人臉特徵辨識方法與系 • 統亦可以改良習知局部二元化圖形(LBP)，在描述人臉影 像之紋理表示時只考慮特徵的存在性(Existence)，而忽略 了其唯一性(Uniqueness)的問題。本發明除了擷取人臉影 像之微觀(Microstructure)紋理資訊外，並進一步採用進階 夕重解析區塊-局部二元化圖形(Advance 犯&如η201205471 VI. Description of the invention: [Technical field to which the invention pertains] The present invention relates to a method and system for identifying biological features, and in particular to a method and system for identifying facial features. [Prior Art] Biometrics technology is playing an increasingly important role in current society. Biometrics applications can be found in applications such as cash machines, access control systems, notebook computers, and portable drives. # In the field of biometrics, face recognition technology is an emerging, research and development focus. Most of the algorithms used in the conventional face recognition system are based on the /"nuclear feature" feature. The so-called "models take advantage of the mathematical dimension reduction model (Dimensi〇n Reducti〇n M〇del) to simulate images. Common methods such as principal component analysis (Principie Component Analysis) or linear identification analysis D1SCriminant Analysis) They are all global (Η_(10) simulation methods. However, this kind of conventional method of simulating global imagery, because it is a feature of taking all the financial images, the material obtained by Lang is not well normalized and some images are partially destroyed (〇 Cdusi〇n) or rotation (r〇 加丨), displacement (shift), size (Scale), and light # (IUuminati〇n) and other damage effects, factors are affected. * These factors will affect the final comparison The feature inward causes the subsequent feature vector comparison to obtain an incorrect identification result. In view of the above, the present invention discloses a face feature recognition method 盥 system 'f is a binary image (L〇cal Binary pattem) Based on the Gaussian Mixture M〇dd model, the feature point distribution information is simulated to solve the problem of familiar face feature recognition. 201205471 One aspect of the present invention is to provide a face feature recognition method. According to the present invention, a face recognition method includes the following steps: (S1) from a video stream. Take a §fl box, (S2) to determine whether there is a face image in the video frame; (S3) if there is the face image, then normalize the face image (N〇rmalize); (S4) utilize one A localized binary image (LBP) is used to describe a set of features of the normalized facial image; and (S5) a Gaussian Mixture Model (GMM) is used to simulate the set of features. Spatially distributed information. In the continuous application, the localized binary pattern used in the face feature recognition method of the present invention is an advanced multiple analytical block-local dualized graph (Advance Multi-res〇). The step (S4) further includes the following sub-steps: (S41) dividing the normalized face image into a plurality of regions centered on a central pixel Block and the plurality of blocks The blocks are arranged in an array, and adjacent blocks have a predetermined distance; (S42) performing a localized binary graph operation by using an average gray scale value of a central block and an average gray scale value of the adjacent block 'by virtue of generating the feature value of the central pixel; and (S43) calculating the feature value of the mother pixel on the face image, thereby generating the set of features. In practical applications, the predetermined distance between the blocks is defined by the center pixel (Center Pixel) and a plurality of surrounding pixels surrounding the central pixel (the Neighbor Pixels), and the method (S4) of the method of the present invention is performed by A predetermined distance between the central pixel and the surrounding pixels is amplified to describe the set of features of the normalized facial image, and the Macrostructure texture information of the facial image is normalized. 201205471 Another aspect of the present invention is to provide a facial feature recognition system. According to another embodiment of the present invention, the present invention provides a facial feature recognition system including a video capture module, a determination module, a conversion module, and an analog module. The image capture module is used to dig a video frame from a stream of silos. The judging module is connected to the image reclining module to determine whether there is a human face image in the video frame, and the face image is normalized if the human face image is present. The conversion module is coupled to the determination module to describe a set of features of the normalized facial image using a partial binary pattern (LBP). The analog module is coupled to the conversion module to simulate a spatial distribution of the set of features using a Gaussian Mixture Model (GMM). In practical applications, the localized binary image used by the facial feature recognition system of the present invention is an advanced multi-resolution block (Local Binary Pattern) (AMB-LBP). And dividing the normalized face image into a plurality of blocks centered on a central pixel, wherein the plurality of blocks are arranged in an array, and adjacent blocks have a predetermined distance therebetween; Then, using the average grayscale value of a central block and the average grayscale value of the adjacent block to perform a localized binarization graph operation to generate the feature value of the central pixel; finally, calculating the facial image The feature value of each pixel to generate the syndrome. In a continuous application, the predetermined distance between the blocks is defined by the center pixel (Center Pixel) and a plurality of surrounding pixels surrounding the central pixel (Neighbor Pixels), and the conversion module of the system of the present invention By amplifying the predetermined distance between the central pixel and the surrounding pixels, the set of features of the face image after the 201205471 image is described and the Macrostructure texture information of the face image is normalized. Compared with the prior art, the face feature recognition method and system of the present invention utilizes a localized binarization graph (LBP) to describe a set of features of the normalized face scene image, and then utilizes Gaussian mixture. The model (GMM) simulates the spatial distribution information of the set of features, thereby constructing features that can resist image damage factors such as shift, rotation, or scale. In addition, the face recognition method and system of the present invention can also improve the conventional local binary pattern (LBP), and only consider the existence of the feature when describing the texture representation of the face image, and ignore the The problem of its uniqueness. In addition to extracting the microstructure texture information of the human face image, the present invention further adopts an advanced analytic block-local binary graph (Advance sin &η;
Block - Local Binary Pattern, AMB-LBP) , 象之巨觀(Macrostmcture)紋理資訊，進而同時兼顧特徵的 f在性與唯—性。綜合上述’本發_較於習知技術即擁 _ 有較佳的辨識率。 所附圖式得到進一步的瞭解。 關於本發明之優點與精神可以藉由町的發明詳述及 【實施方式】 本發明之一範疇在於提供 1 0。諸奋 HB 向 *4- /.Λν - . 。凊參閱圖一，其繪示根據本發明之 臉特徵辨識方法10之流程圖。 種人臉特徵辨識方法 明之一具體實施例的人 如圖一所示，本發明人臉特徵辨識方法1G包含有以 201205471 下步驟：（si)自一視訊串流中擷取一視訊框；（S2)判斷該 視訊框中是否有一人臉影像；（S3)若有該人臉影像，則將 故八臉彰像正規化(Normalize) ; (S4)利用一局部二元化圖 形(Local Binary Pattern, LBP)來描述正規化後該人臉影像 之、且特徵，（S5)利用一高斯混合模型（Gaussian Mixture Model，GMM)來模擬該組特徵之空間分布資訊。 «月參閱圖二，圖二係繪示本發明之一具體實施例的進 Pfc白多重解析區塊-局部二元化圖形。於實際應用中，本發 明人臉特徵辨識方法1〇所採用之局部二元化圖形，係為 一種進階多重解析區塊-局部二元化圖形(Advance Muki_ resolution Block - Local Binary Pattern，AMB-LBP)，而步 驟(S4)則進-步包含有町子步驟：（S41)將經正規化後之 該人臉影像’以-中央像素c為中心、分割成複數個區塊 32 ’而該等複數個區塊32係⑽列式排列，且相鄰之該 區塊32間係具有一預定距離d ; (S42)利用一中央區塊之 平均灰階值與其相鄰區塊之平均灰階值，進行局部二元化 圖形運算，以產生該巾央像素c之特徵值；以及(S43)計 算該人臉影像上之每-像素之特徵值，以產生該組特徵。 於實際應用中，該區塊32間之預定距離d係利用該 中心像素(Center Pixel)C以及8個圍繞該中心像素c之周Λ 圍像素(Neighbor Pixels)0小2 7來定義，而本發明方 法10之步驟(S4)係藉由擴增該中心像素c與該周圍像素 〇小2...7之預定距離d，來描述該經正規化後人臉影像 201205471 之該組特徵，並取得經正規化後 ,A/r 设的該人臉影像之巨觀 (Macrostructure)紋理資訊。 請參閱圖-。本發明人臉特徵 以下步驟：（S6)利用一多類支括° 3有 c , Λ7 叉待向量機（Multi_Class 針對該組賴之空布資訊， 構-辑產生錢H ;叹(S7胸肋特徵之空 :刀布”fl ’而利用一多類支持向量機__咖Block - Local Binary Pattern, AMB-LBP), Macrostmcture texture information, and at the same time take into account the characteristics of the characteristics of f and sexuality. The above-mentioned 'this is a better recognition rate than the conventional technology. The drawings are further understood. The advantages and spirit of the present invention can be explained by the invention of the town and [Embodiment] One aspect of the present invention is to provide 10. Zhufen HB to *4- /.Λν - . Referring to Figure 1, there is shown a flow chart of a face feature recognition method 10 in accordance with the present invention. As shown in FIG. 1 , the face feature recognition method 1G of the present invention includes the steps of 201205471: (si) capturing a video frame from a video stream; S2) determining whether there is a face image in the video frame; (S3) if there is the face image, normalizing the eight face image (Normalize); (S4) using a local binary pattern (Local Binary Pattern) (LBP) to describe the features and features of the face image after normalization, (S5) using a Gaussian Mixture Model (GMM) to simulate the spatial distribution information of the set of features. «Month Referring to FIG. 2, FIG. 2 is a diagram showing a Pfc white multi-resolution block-local binarization pattern according to an embodiment of the present invention. In practical applications, the local binary feature used in the face feature recognition method of the present invention is an advanced multi-resolution block-local Binary Pattern (AMB-). LBP), and the step (S4) further includes a step of the town: (S41) dividing the normalized face image 'centered on the central pixel c into a plurality of blocks 32' And a plurality of blocks 32 are arranged in a column (10), and adjacent blocks 32 have a predetermined distance d; (S42) using an average gray level value of a central block and an average gray level of the adjacent block a value, performing a localized binary graphics operation to generate a feature value of the towel center pixel c; and (S43) calculating a feature value of each pixel on the face image to generate the set of features. In practical applications, the predetermined distance d between the blocks 32 is defined by the center pixel (C Center C) and eight surrounding pixels (Neighbor Pixels) of the central pixel c. The step (S4) of the method of the invention 10 describes the set of features of the normalized face image 201205471 by amplifying the predetermined distance d between the central pixel c and the surrounding pixels by 2...7, and After the normalization, A/r sets the Macrostructure texture information of the face image. Please refer to the figure -. The following steps are performed on the face feature of the present invention: (S6) utilizes a plurality of types of support, and has a c, Λ7 forked vector machine (Multi_Class for the group of the empty cloth information, the structure-set generates money H; sighs (S7 chest ribs) Feature space: knife cloth "fl" and use a multi-class support vector machine __ coffee
Support Vector Machines)來產生一人臉辨識結果。 /參關…於實際應时，本發明之人臉特徵辨識 / 10可以刀為Di|練階段（Training)以及測試階段 (Testing) ’ 訓練階段藉由步驟(S1)、（S2)、（s3)、（s4)、（s5) 以及(S6)來建構訓練產生分類器；測試階段則藉由所建構 完成之訓練產生分類器12與該等步驟(si)、（s2)、（s3)、 (S4)、（S5)以及(S7)來產生人臉辨識結果。 相較於習知技術，本發明之人臉特徵辨識方法1〇可 以利用局部二7〇化圖形(LBP)，來贿正規化後人臉影像 之一組特徵，接著利用高斯混合模型（GMM)來模擬該組 特徵的空間分布資訊’以建構出可以抵抗偏移(Shift)、些 微旋轉(Rotation)、或是尺寸(scaie)變化等影像破壞因素的 特徵。此外，本發明之人臉特徵辨識方法10亦可以改良 習知局部二元化圖形(LBP) ’在描述人臉影像之紋理表示 時，只考慮特徵的存在性(Existence)，而忽略了其唯一性 (Uniqueness)的問題。本發明除了擷取人臉影像之微觀 (Microstructure)紋理資訊外，並進一步採用進階多重解析 £ 塊-局部一元化圖形（Advance Multi-resolution Block-Local Binary Pattern, AMB-LBP) ， 以擷取人臉影像之巨觀 201205471 (Macr〇stnicture)紋理資訊’進而同日夺兼顧特徵的存在性盘 :綜合上述，本發明相較於習知技術即擁有較佳的 y, _1 5E\j —j— 本毛月之另範嚀在於提供一種人臉特徵辨識系統 ^0 ° 請參閱圖三，圖三係綠示本發明之一具體實施例的人 臉特徵辨識系統20之功能方塊圖。根據本發明之另一具 體實施例’本發明所提供之—種人麟徵賴系統2〇包 含有一影像擷取模組22、一判斷模組24、一轉換模組％ 以及一模擬模組28。影像擷取模組22係用以自一視訊串 流中擷取一視訊框。判斷模組24係與影像擷取模組22連 接，以判斷視訊框中是否有一人臉影像，若其中有人臉影 像，則將人臉影像正規化(N〇rmaiize)。轉換模組％係連 接於判斷模組24，以利用一局部二元化圖形(L〇cal Binary Pattern，LBP)，來描述經正規化後人臉影像之一組特徵。 模擬模組28係與轉換模組26連接，以利用一高斯混合模 型（Gaussian Mixture Model，GMM)來模擬該組特徵之空間 分布資訊。 請參閱圖二。於實際應用中，本發明人臉特徵辨識系 統20所採用之局部二元化圖形，係為一種進階多重解析 區塊-局部二元化圖形（Advance Multi-resolution Block-Local Binary Pattern， AMB-LBP) ’ 其係將經正規化後之該 人臉影像’以一中央像素C為中心分割成複數個區塊 32(如圖二所示灰色區域之區塊），而該複數個區塊32係 以陣列式排列，且相鄰之該區塊32間具有一預定距離 d;接著’利用一中央區塊之平均灰階值與其相鄰區塊之 201205471 平均灰階值，進行局部二元化圖形運算，以產生中 c之特徵值，最後，計算該人臉影像上母一像素之特^支 值，以產生該組特徵。 *Support Vector Machines) to generate a face recognition result. / Participation... In actual time, the face feature recognition/10 of the present invention can be a Di|training phase and a testing phase. The training phase is performed by steps (S1), (S2), (s3). , (s4), (s5), and (S6) to construct a training-generating classifier; in the test phase, the classifier 12 is generated by the constructed training and the steps (si), (s2), (s3), (S4), (S5), and (S7) generate face recognition results. Compared with the prior art, the face feature recognition method of the present invention can utilize a local two-dimensional graph (LBP) to bribe a group of facial image features, and then utilize a Gaussian mixture model (GMM). To simulate the spatial distribution information of the set of features' to construct features that can resist image damage factors such as offset (Shift), micro-rotation (Rotation), or scaie variation. In addition, the face feature recognition method 10 of the present invention can also improve the conventional localized binary pattern (LBP). When describing the texture representation of the face image, only the existence of the feature is considered, and the uniqueness is ignored. The problem of uniqueness. In addition to extracting the microstructure texture information of the face image, the present invention further adopts the Advance Multi-resolution Block-Local Binary Pattern (AMB-LBP) to capture the person. The image of the face image 201205471 (Macr〇stnicture) texture information 'and then the same day and the characteristics of the existence of the disk: In summary, the present invention has better y, _1 5E\j -j - this book compared to the prior art Another example of Maoyue is to provide a facial feature recognition system. Referring to FIG. 3, FIG. 3 is a functional block diagram of the facial feature recognition system 20 of one embodiment of the present invention. According to another embodiment of the present invention, the present invention provides an image capture module 22, a determination module 24, a conversion module %, and an analog module 28. . The image capture module 22 is configured to capture a video frame from a video stream. The judging module 24 is connected to the image capturing module 22 to determine whether there is a human face image in the video frame, and if there is a human face image, the facial image is normalized (N〇rmaiize). The conversion module % is connected to the determination module 24 to describe a group feature of the normalized face image by using a partial binary pattern (LBP). The analog module 28 is coupled to the conversion module 26 to simulate the spatial distribution of the set of features using a Gaussian Mixture Model (GMM). Please refer to Figure 2. In practical applications, the local binary pattern used by the facial feature recognition system 20 of the present invention is an advanced multi-resolution block-local Binary Pattern (AMB-). LBP) 'The face image of the normalized face is divided into a plurality of blocks 32 (blocks of the gray area as shown in FIG. 2) centered on a central pixel C, and the plurality of blocks 32 Arranged in an array, and adjacent blocks 32 have a predetermined distance d; then 'localized dualization using the average gray level value of a central block and the 201205471 average gray level value of its adjacent block The graphics operation is performed to generate the feature value of the medium c. Finally, the special value of the parent pixel on the face image is calculated to generate the group feature. *
於實際應用中，該區塊32間之預定距離d係利用— 中心像素(Center Pixel)C，以及8個圍繞該中心像素c之 周圍像素(Neighbor Pixels)0、1、2".7來定義，而本發明 系統20之轉換模組26係藉由擴增該中心像素c與該周 圍像素0、1、2...7之預定距離d，來描述該經正規化後 的人臉影像之該組特徵，並取得經正規化後該人臉影像之 巨觀(Macrostructure)紋理資訊。 請參閱圖三，本發明人臉特徵辨識系統2〇另包含有 一分類辨識模組30，其係與模擬模組28連接，而利用一 多類支持向量機(Multi-Class Support Vector Machines)針對 該組特徵之空間分布資訊，來建構一訓練產生分類器l2 或產生一人臉辨識結果。 ^請參閱圖三，於實際應用中，本發明之人臉特徵辨識 系統20可以分為訓練階段（Training)以及測試階段 (Testing)。訓練階段係藉由影像擷取模組22、判斷模組 24、轉換模組26、模擬模組28以及分類辨識模組3〇來 建構训練產生分類器12。測試階段則係藉由建構完成之 麟產生分類H 12以及影像#模組22、判斷模組24、 轉換模組26、模擬模、组28與分類辨識模組3〇來產生人 臉辨識結果。 201205471 人臉影像之徵’接著模擬模組π會湘高斯混合 杈型（GMM)來模擬該組特徵的空間分布資訊，以建構出 可以抵抗滿移（shift)，些微旋轉（R〇加i〇n)、或是尺寸 (Scale)變化等影像破壞因素的特徵。此外’本發明人臉特 徵辨識系統20之轉換模組26，亦可以改良習知局部二元 化圖形(LBP)於描述人臉影像之紋理表示時，只考慮特徵 ，存在性(Existence) ’而忽略了其唯一性(Uniqueness)的問 題。本發明除了操取人臉影像之微觀(Micr〇structure)紋理 貧訊外，並進一步採用進階多重解析區塊_局部二元化圖 (Advance Multi-resolution Block - Local Binary Pattern, AMB LBP) ’以擷取人臉影像之巨觀(Macr〇structure)紋理 資訊’進而同時兼顧特徵的存在性與唯一性。綜合上述’ 本發明相較於習知技術即擁有較佳的辨識率。 藉由以上較佳具體實施例之詳述，係希望能更加清 楚描述本發明之特徵與精神’而並非以上述所揭露的較佳 具體實施例來對本發明之範疇加以限制。相反地，其目的 疋希望月b/函盖各種改變及具相等性的安排於本發明所欲申 請之專利範圍的範缚内。 【圖式簡單說明】 圖一繪示根據本發明之一具體實施例的人臉特徵辨識 方法之流程圖。 圖二係繪示本發明之一具體實施例的進階多重解析區 塊-局部二元化圖形。 圖三係繪示本發明之一具體實施例的人臉特徵辨識系 統之功能方塊圖。 201205471 【主要元件符號說明】 C ： 中心像素 0、 1、2....7 :周圍像素 in · λ. w - 人臉特徵锊識方法 λ 1Z :訓竦產生分類器 20 ： 人臉特徵辨識系統 22 :影像擷取模組 24 ： 判斷模組 26 :轉換模組 28 ： 模擬模組 30 :分類辨識模組 32 ： 區塊 S1〜S7 :實施步驟In practical applications, the predetermined distance d between the blocks 32 is defined by - Center Pixel C, and 8 surrounding pixels (Neighbor Pixels) 0, 1, 2 " The conversion module 26 of the system 20 of the present invention describes the normalized facial image by amplifying the predetermined distance d between the central pixel c and the surrounding pixels 0, 1, 2, . . . The set of features, and the Macrostructure texture information of the face image after normalization is obtained. Referring to FIG. 3, the facial feature recognition system 2 of the present invention further includes a classification identification module 30 connected to the simulation module 28 and using a multi-class support vector machine (Multi-Class Support Vector Machines). The spatial distribution information of the group features is used to construct a training to generate the classifier l2 or generate a face recognition result. Referring to FIG. 3, in practical applications, the facial feature recognition system 20 of the present invention can be divided into a training phase and a testing phase. In the training phase, the training generation classifier 12 is constructed by the image capturing module 22, the determining module 24, the converting module 26, the simulation module 28, and the classification recognition module 3A. In the test phase, the face recognition result is generated by the construction of the classification H 12 and the image #module 22, the determination module 24, the conversion module 26, the simulation module, the group 28 and the classification recognition module 3〇. 201205471 The phenomenon of facial image 'then simulates the module π will be Gaussian mixed 杈 type (GMM) to simulate the spatial distribution information of the set of features, to construct a resistance to full shift (shift), some micro-rotation (R〇 plus i〇 n), or the characteristics of image destruction factors such as size changes. In addition, the conversion module 26 of the facial feature recognition system 20 of the present invention can also improve the conventional partial binary graphics (LBP) to describe the texture representation of the face image, and only consider the feature, existence (Existence) Ignore the problem of its uniqueness. In addition to the microscopic (Micr〇structure) texture poorness of the face image, the present invention further adopts the Advance Multi-resolution Block (Local Binary Pattern (AMB LBP)'. In order to capture the macroscopic (Macr〇structure) texture information of the face image, and at the same time take into account the existence and uniqueness of the feature. The above-mentioned invention has a better recognition rate than the prior art. The scope of the present invention is limited by the preferred embodiments of the invention disclosed herein. On the contrary, it is intended that the various changes and equivalents of the month b/ letter are within the scope of the patent scope of the invention as claimed. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flow chart showing a face feature recognition method according to an embodiment of the present invention. Figure 2 is a diagram showing an advanced multi-resolution block-local binarization pattern in accordance with an embodiment of the present invention. Figure 3 is a functional block diagram of a face feature recognition system in accordance with an embodiment of the present invention. 201205471 [Description of main component symbols] C : Center pixel 0, 1, 2, ....7: Surrounding pixels in · λ. w - Face feature recognition method λ 1Z : Training generation classifier 20 : Face feature recognition System 22: image capture module 24: determination module 26: conversion module 28: analog module 30: classification recognition module 32: blocks S1 to S7: implementation steps
i s] 12i s] 12