201223505 六、發明說明： 【發明所屬之技術領域】 本發明係關於一種醫學分析方法及醫學分析裝置，特 別關於一種使用非接觸式之呼吸量測之醫學分析方去及 醫學分析裝置。 / 【先前技術】 最近，市面上出現一些可用以呼吸量測的醫療裴置， 其包含複數穿戴件，需讓使用者穿戴在頭上或胸部，利用 呼吸的起伏來測量呼吸的狀況，可協助判斷受測者是否有 打軒、或呼吸中止症等等、或者是用來協助重症病人觀測 其呼吸是否正常等等。 、由於上述量測呼吸的設備，皆須接觸使用者，因而造 成使用者在錢上的不舒適感，甚至嚴重影f生活與睡眠 =質’也在某種程度上限制了呼吸量測的使用時機及❹ 場合。 因此，如何提供一種醫學分析方法及醫學分析裝置， 能夠在不干擾受測者的情況下，提供高正確性的呼吸量測 資訊以作出廣泛及精確的醫學分析，實為當前重要課題之 ——0 【發明内容】 、有鐘於上述課題，本發明之目的為提供-種醫學分析 方法及醫學分析裝M，能夠在不谓受測者的情況下，提 201223505 供高正確性的啤吸量測資訊 析。 以作出廣泛及精確的醫學分 述目的’依據本發明之—種藉由呼吸量測而進 子諸方法包括下列步驟一影像擷取㈣ 複數影像’·—狀1判斷步驟，依據該等影像判斷 3亥生物位於—可量職態或—不可量難態； 當該生物位於該可量測狀態時，進行該生物之料 量m分析步驟’依據該料量測結 況進行分析。 狀 磐j 目的’本發㈣露—種具有呼吸量測功能之 子刀斤裝置’其包括一影像擷取模組、一狀態判斷模 且、-料量_組以及—分析模組。歸獅模組對一、 物掏取複數影像。狀態判斷模、组依據該等影像判斷該生 於-可量測狀態或一不可量測狀態。呼吸量測模組係 虽该生物位於該可量測狀態時，進行該生物之呼吸量測。 分析模組㈣該呼吸量測模組之呼吸㈣結果對 狀況進行分析。 切 八在：實施例中，狀態判斷步驟包含：依據該等影像而 刀析出影像之-前景或―背景；依據該等影像擷取該前景 之至少-面積特徵；以及將該面積特徵放入一判斷模型而 判斷該生物位於該可量測狀態或該不可量測狀態。藉由判 斷該生物位於可4測狀態或該不可量測狀態，可大幅提言 呼吸量測資訊的正確性及可信賴度。 门 在-實施例中’狀態判斷步驟包含：依據該等影像而 201223505 刀斤出影像之一前景或一背景；依據該等影像擷取該前景 之至少-面積特徵；以及比對該面積特徵與一闕值，而判 斷5亥生物位於該可量測狀態或該不可量測狀態。除了上述 利用判斷模型來進行狀態判斷步驟之外，本發明亦可利用 較易之判斷方法，例如比對該面積特徵與一閥值 生物之量測狀態。 m f一實施例中，面積特徵包括單一影像之前景之面 積、前後影像之前景之面積之變化量、或前後影像之前景 之互斥面積的變化量。藉由不同的面積特徵可達到不同的 比對的效果，並可應用於不同量測場合。 在一實施例中，於狀態判斷步驟中，更判斷該生物位 於介於可量測狀態與不可量測狀態之間之一轉換狀態。藉 由增加之轉換狀態，可提升狀態判斷的效能，進而提升^ 吸量測的正確性。 在一實施例中，一收斂系數係隨生物位於可量測狀 態、不可量測狀態、或轉換狀態而改變大小以控制可量測 狀態、不可量測狀態、.或轉換狀態之收斂速度。例如，當 生物位於可量測狀態時，收斂系數較小，此時收斂速度^ 慢，以使判斷模型對變化較為不敏感；當生物位於不可量 測狀態時，收斂系數為零，此時收4速度趨近於零或等於 零；當該生物位於轉換狀態時，該收斂系數較大，此時收 斂速度較快，以使判斷模型對變化較為敏感。藉由收斂系 數可提高狀態判斷的速度及效能。 在一實施例中，呼吸量測步驟包含：依據該等影像判 201223505 _生物之-呼吸方向’·以及依據該等影像產生該呼吸方 向，呼吸曲線。本發明可適用於使用者仰躺、側躺、站 立等不同的姿勢，對應這些不同的姿勢，生物的呼吸方向 會不同，故本發明藉由先判斷生物之呼吸方向，再產生該 呼吸方向之呼吸曲線，進而大幅提升呼吸量測資訊的準確 性。 在-實施例中，分析步驟係對該生物分析為睡眠呼吸 中止症、淑死病人臨終判斷、氣喘、幼兒啤吸暫停、過渡 •換氣症候群、成人呼吸箸迫症、肺栓塞、急性肺水腫、心 :肺停止、或嗎啡中毒及其他與呼吸具相關連之症狀。 • 纟上所述’本發明之醫學分析方法及醫學分析裝置係 採用非接觸式的呼吸量測方式，藉由對生物進行影像擷 取，並藉由分析影像而得到該生物之呼吸狀況，進而利用 啤吸狀況分析該生物狀況。本套人性化智慧方法乃使用非 接觸式之視sfL影像分析技術，用以量測使用者的呼吸資 身訊。使用者不需改變任何生活習慣與擔心儀器影響活動自 由纟統即會自動分斩使用者的呼吸情境，進而動態與精 準的量測使用者之呼吸資訊。此外，本發明在呼吸量測之 前’會先判斷該生物位於可量測狀態（例如靜止或小幅動 乍）或疋不可量測狀態（例如大幅動作），以致所得到的 呼吸量測資訊之可靠度及精確度皆大幅提升，致使本發明 可應用更廣泛及精確的病症分析。此方法及裝置不僅適人 =家和遠距照護’更可以提供安養機構和醫療中心做; 吸為Λ量測，以做病症和健康的輔助判斷與監控。 201223505 【實施方式】 以下將參照相關圖式，說明依本發明較佳實施例之一 種醫學分析方法及醫學分析裝置，其中相同的元件將以相 同的參照符號加以說明。 圖為本發明較佳實施例之一種醫學分析方法的流程 圆立圖A本發明較佳實施例之一種醫學分析裝置的方塊 不意圖，圖3為本發明較佳實施例之醫學分析方法及醫學 分析裝置應用的情境示意圖。 明’’居圖1至圖3以說明本發明較佳實施例之醫學分 析方法及，分析t置。本實關之醫學分析方法包含一 影像擷取步驟S〇卜一狀態判斷步驟S02、-呼吸量測步 及-分析步驟SG4。本實施例之醫學分析裝置1〇 匕3衫像擷取模組U、-狀態判斷模組12、-呼吸量 二、’ 13卩及”析模、址14。本實施例所應用的情境係 =病的小孩進行非接觸式的呼吸量測並提供小孩狀況 刀析，當然，这僅為舉例說明，並非用以限制本發明。 =，本實施例之狀態判斷模組12、呼吸量測模組⑴乂 及为析模組14係整合為—電腦系统c作例說明。 首先，在影像擷取步，驟S01中係藉由影像摘取模 對-生物擷取複數影像。影_取模組u可包含 析度的攝影機以提供高解析度的影像。由圖3可知，^ =凡全不需要對f測者進行接觸，所以受測者 制，而可安穩的休息。 又限 接著，狀態判斷步驟S02係藉由狀態判斷模組η依 201223505 據該等影像判斷該生物位於一可量測狀態或一不可量測 狀態。於此，所謂可量測狀態可包含生物完全靜止或小幅 動作，例如手腳的小幅動作等不影響呼吸量測之動作。而 不可量測狀態包含生物大幅度的動作等影響呼吸量測之 動作。 在本實施例中，狀態判斷步驟可包含：依據該等影像 而分析出影像之一前景（foreground )或—背^景 (background );依據該等影像擷取該前景之至少—面積特 籲徵；以及將該面積特徵放入一判斷模型而判斷該生物位於 ： 可量測狀態或不可量測狀態。 ' , 在本實施例中，是藉由高斯分佈來模型化背景，並可 據此分離出前景。前景或背景的分析係根據所擷取影像之 畫素（pncel)，由於此部分屬先前技術之應用，故細 再贅述。 、 在分析出前景之後，依據該等影像擷取該前景之至少 /一面積特徵。面積特徵可包括單—影像之前景之面積、二 後影像之前景之面積之變化量、或前後影 :積;變化量。其中，單-影像之前景之面積之 在連續的影像中可反映出前景的變化。前後影像之前 景之面積之變化量之面積特徵，亦可反映出前景的變化， ^變化量越來越大，反映出該生物可能在動作中，例如翻 田變化量越來越小，反映出該生物可能在趨於靜止，’ =如快完成翻身動作。前後影像之前景之互斥面積 量之面積特徵亦可反映出前景的變化，當互斥面積的變化 201223505 量走1來越〗、，表示刖後影像的前景幾近重疊，反映出生物 可育匕趨於靜止，反之，當互斥面積的變化量越來越大，反 映出生物正大幅動作。據此’上述面積特徵皆可用來描述 前景的動作狀態。 田面積特徵求出之後，便可將面積特徵放入一判斷模 型而判斷該生物位於可量測狀態或不可量測狀態。判斷模 型例如為人工智慧模型，如貝氏分類器、類神經網路、 adab〇〇St分類器、SVM分類器、LDA線性判別分析（Fisheris —a1，ClaSSifie〇、0,tsu 分類法、決策樹、K-nearest neighbor (KNN)隱馬爾可夫模型（hiddenmark〇vm〇dei)等等具 判斷能力之方法。藉由判斷該生物位於可量測狀態或該不 可量測狀態，可大幅提高呼吸量測資訊的正確性及可信賴 度。 此外，在狀態判斷步驟中，t可判斷該生物位於介於 可量測狀態與不可量測狀態之間之-轉換狀態。圖4係顯 示本實施例所使用之三種狀態的關係：可量測狀態W、不 可量測狀態S2、轉換狀態S3。且轉換狀態幻係存在於不 可量測狀態S2到可量測狀態S1之間，亦即當生物由大幅 動作（不可量測狀態S2)經過轉換狀態S3而至靜止或小 巾田動作（可量測狀態S1)、或是由轉換狀態S3變換至不可 量測狀態S2。藉由增加之轉換狀態S3，可提升狀態判斷 的效能，進而提升呼吸量測的正確性。 此外’本發明係在判斷模型中增加一收敛系數，其係 隨生物位於可量測狀態、不可量測狀態、或轉換狀態而改 201223505 變大小以控制可量測狀態、不可量測狀態、或轉換狀態之 收斂速度。收斂系數越小，則狀態判斷越敏感，即越容易 感受到變化，此時收斂速度較慢；收斂系數越大，則狀態 判斷越不敏感，即越不容易感受到變化，此時收斂速度較 快。在本實施例中，收斂系數為背景模型參數，用來控制 者景模型之收斂速度，其應用情形可例如如下所述：當系 統處於可量測狀態si (靜止）時，可設收斂系數較小，如 此可使系統不容易受到變化，才有助於監測到生物的動 •作；當系統變為不可量測狀態S2 (大幅動作）時，可設收 r傲系數為零’如此可使系統不因大幅動作受到變化；當系 :統變為轉換狀態S3時，代表生物已慢慢靜止，此時可調 局收斂系數以加速收敛’直到收敛完畢，系統回到可量測 狀態S1，此時又可降低收料數，助於持續監測生物的動 作。藉由收斂系數可提高狀態判斷的速度及效能。 除了上述利用判斷模型來進行狀態判斷步驟之外，本 利用較簡易之判斷方法。在此態樣中，咖 步脉包含：依據該等影像而分析出影像之—前景或一背 景，依據該等影像擷取該前景之至少一面積特徵；以及比 對該面積特徵與一閥值， 或該不可量測狀態。其生物位於該可量測狀態 乂县 具應用_例如如下所述，並以單一 二：景之面積作為面積特徵：若 S】，則面積特徵㈣會趨近於 Θ測狀態 背景）。當使用者開始動時，面=為影像全部幾近就是 „ 面積特徵的值會變大「阳故 前景逐漸出現），因此面積 7值會文大（因為 積特徵右大於某個閥值，系統便 201223505 得知使用者目刚正在移動，而轉換到不可量測狀態s2，並 且設收斂系數為零。當使用者動完靜止時 ，面積特徵會一 直保持某個值（a是因為前景沒有變動），若這個值超過 某段時間皆無太多變化，則代表使用者已經靜止，此時系 統狀態轉到轉換狀態S3，並且調高收敛系數值，以加速收 敛直到收激凡畢，系統會回到可量測狀態si ’並調低收 斂系數值。 另外，本發明亦可使㈣他比對的方式來得知生物是 位於可量測狀態或不可量測狀態。例如，在狀態判斷步驟 中，可不必分析出影像之前景或背景，而可包含：計算生 物移動之移動動量，例如移動向量（m。—ve咖）；以及 :用-個決策機制判斷該移動動量是否符合呼吸量測的 條件’即判斷生物位於可量測狀態或不可量測狀態。由於 移動向量係屬習知技術，故其細節於此不再費述。決策機 ::例如將移動動量與—閥值相比較、或是先得出移動動 量隨時間的變化，再賴變化與—解、―週期 幅相比較。 派 進態時’可_吸量測模組 該等影像包含:依據 —以及依據該等影像產生該呼吸;向二= 線。本發明可適用於使用者仰躺、侧躺、站 勢，對應這些不同的姿勢，生物的如及5的姿 發明藉由先判斷生物之呼吸方向，再產生該呼=向 12 201223505 吸曲線，進而大幅提升呼吸量測資訊的準確性。 以下舉例說明呼吸方向及呼吸曲線的求得過程。 對某一段含有呼吸之連續影像作光流（〇ptical fl〇w )， 以得到兩兩影像間之移動向量（motion vector)，記作 MU，其中，X與y代表晝素的座標，j代表第』個影像。 此外κ分作水平與垂直的分量，分別記作册（)與 ⑽）。再以下列方程式分別累加·^與⑽也)所示： ϋ鑛^ | (x,y) ΣΣ- ^ f及代曲線如圖5所示。其中，H代表畫素之y座 仏的極限值，w代表晝素之u標的極限值。 為決mm作為呼吸方向，可算出它們的變異數 二記作V^vv。當vhA於”，則 方向作動呼吸方向，當Vh τ 呼吸方向。當則以垂直方向作動 決定…&方向時，記作^於此係以 7乍為為例，矸的曲線如圖6所示 曲線=!吸:線’_時間轴上:累加(記作 所示）的動作： 。己作⑽7 ’曲線如圖7 W，其中W設為零。 &4S；= 13 201223505 以目前的點，並 作為平滑的判斷 其中m為平滑視窗的大小，意思為： 且往前和往後各取m點，共2m+l個點， 依據。 或需注意者，上述將呼吸方向拆分為水平及垂直方向僅 錄制本發明，本發明料因情況不 同而疋義不同及更多的呼吸方向。 在求得呼吸曲線之後，更可進行分析步驟sq4，其係 藉由分析模組Μ依據該呼吸量測結果對該生物狀況進行 分析，分析可包含算出呼吸特徵、统計、製作成報告或病 狀可能性預測、或診斷。例如對該生物分析為睡眠呼吸中 止症、瀕死病人臨終判斷、氣喘、幼兒呼吸暫停、過渡換 氣症候群、成人呼吸箸迫症、肺栓塞、急性肺水腫、:肺 停止、或嗎啡中毒及其他與呼吸具相關連之症狀。上述症 狀表現在呼吸的情況如下所述：睡眠呼吸中业症之初步判 斷（胸部與腹部之呼吸會不一致，而且會有呼吸停止之特 性）；瀕死病人臨終判斷（呼吸的深度和兩兩呼吸的時間 間隔異常）；氣喘（呼吸困難）；幼兒呼吸暫停而猝死（呼 吸停止）；過渡換氣症候群（呼吸過快、過深）；成人呼吸 窘迫症（呼吸深且快）；肺栓塞（呼吸加重，頻率加快）； 心丨生肺水腫（呼吸短促）；心肺停止（呼吸速度下降）丨嗎 啡中毒（呼吸速度下降）。本發明可藉由呼吸曲線所得到 的資料，其中至少包含呼吸強度（振幅）、頻率、時間、 走勢等等，或者佐以其他輔助資料來得到精確的判斷。 綜上所述，本發明之醫學分析方法及醫學分析裝置係 14201223505 VI. Description of the Invention: [Technical Field] The present invention relates to a medical analysis method and a medical analysis device, and more particularly to a medical analysis method and a medical analysis device using a non-contact respiratory measurement. / [Prior Art] Recently, there are some medical devices available for respiratory measurement, which include a plurality of wearing parts, which are required to be worn by the user on the head or chest, and the breathing condition is used to measure the breathing condition, which can assist in judging Whether the subject has a diarrhea, or a respiratory arrest, etc., or is used to assist a critically ill patient to observe whether his breathing is normal or the like. Because the above-mentioned equipment for measuring breathing has to contact the user, it causes the user's discomfort in money, and even seriously affects the life and sleep = quality also limits the use of respiratory measurement to some extent. Timing and occasions. Therefore, how to provide a medical analysis method and a medical analysis device, which can provide high-precision respiratory measurement information without disturbing the subject to make extensive and accurate medical analysis, is an important issue at present - 0 [Summary of the Invention] The present invention aims to provide a medical analysis method and a medical analysis device M, which can provide 201223505 for high correctness of beer suction without the subject being tested. Information analysis. For the purpose of making a broad and precise medical description, the method according to the present invention includes the following steps: image capture (4) complex image '·-form 1 judgment step, based on the images The 3H bio is located in a measurable position or a non-quantitative state; when the organism is in the measurable state, the step of performing the mass m analysis of the organism is analyzed according to the measured condition. The object of the present invention is a sub-knife device having a respiratory measurement function, which includes an image capturing module, a state judging module, a material amount group, and an analysis module. The lion module is used to capture multiple images of the object. The state judging mode and the group judge whether the state is in a measurable state or an unmeasurable state based on the images. The spirometry module performs a respiratory measurement of the creature while the creature is in the measurable state. The analysis module (4) The breathing (4) results of the respiratory measurement module analyze the condition. In the embodiment, the state determining step includes: separating the foreground- or foreground-image of the image according to the images; extracting at least the area feature of the foreground according to the images; and placing the area feature into the image The model is judged to determine that the creature is in the measurable state or the unmeasurable state. By judging that the creature is in a measurable state or in an unmeasurable state, the correctness and reliability of the spirometry information can be greatly enhanced. In the embodiment, the state determination step includes: according to the images, 201223505 punches out a foreground or a background of the image; extracts at least an area characteristic of the foreground according to the images; and compares the area characteristics with the image A value is determined, and it is judged that the 5H creature is located in the measurable state or the unmeasurable state. In addition to the above-described state determination step using the judgment model, the present invention can also utilize a relatively easy judgment method, such as a measurement state of the area feature and a threshold value. In an embodiment, the area feature includes the area of the front view of the single image, the amount of change in the area of the front and back of the front and back images, or the amount of change in the mutually exclusive area of the front and back images. Different contrast characteristics can be achieved by different area features, and can be applied to different measurement occasions. In an embodiment, in the state determining step, the biometric is further determined to be in a transition state between the measurable state and the unmeasurable state. By increasing the transition state, the performance of the state judgment can be improved, thereby improving the correctness of the measurement. In one embodiment, a convergence coefficient is varied as the organism is in a measurable state, a non-measurable state, or a transition state to control the convergence rate of the measurable state, the unmeasurable state, or the transition state. For example, when the creature is in the measurable state, the convergence coefficient is small, and the convergence speed is slow, so that the judgment model is less sensitive to the change; when the creature is in the unmeasurable state, the convergence coefficient is zero, and the convergence coefficient is 4 The speed approaches zero or equals zero; when the creature is in the transition state, the convergence coefficient is larger, and the convergence speed is faster, so that the judgment model is more sensitive to the change. The speed and performance of state judgment can be improved by the convergence factor. In one embodiment, the respiratory measurement step includes: determining the respiratory direction and the breathing curve based on the images according to the images 201223505 _ bio-respiratory direction'. The invention can be applied to different postures such as lying on the back, lying on the side, standing, etc., and the breathing direction of the creature is different according to the different postures, so the invention first determines the breathing direction of the creature, and then generates the breathing direction. The breathing curve, which in turn greatly improves the accuracy of the respiratory measurement information. In an embodiment, the analyzing step is the biological analysis of sleep apnea, dying of a dying patient, asthma, palpation of a child, transitional/ventilation syndrome, adult respiratory distress, pulmonary embolism, acute pulmonary edema Heart: lungs stop, or morphine poisoning and other symptoms associated with breathing apparatus. • The above-mentioned medical analysis method and medical analysis device of the present invention adopts a non-contact respiratory measurement method, and obtains the respiratory state of the living body by analyzing the image and then obtaining the respiratory state of the living body. The biological condition was analyzed using the condition of the beer. This set of humanized wisdom methods uses non-contact sfL image analysis technology to measure the user's respiratory intelligence. Users do not need to change any living habits and worry about the instrument's influence. The activity system automatically separates the user's breathing situation, and then dynamically and accurately measures the user's breathing information. In addition, the present invention will first determine that the organism is in a measurable state (eg, static or small motion) or an unmeasurable state (eg, a large motion) before the respiratory measurement, so that the obtained respiratory measurement information is reliable. Both the degree and accuracy are greatly enhanced, resulting in the application of a broader and more accurate analysis of the condition of the present invention. This method and device is not only suitable for people and home care, but also provides security institutions and medical centers to do; as a measure of sputum, to make judgments and monitoring of illness and health. [Embodiment] Hereinafter, a medical analysis method and a medical analysis apparatus according to a preferred embodiment of the present invention will be described with reference to the accompanying drawings, wherein the same elements will be described with the same reference numerals. The figure is a flow chart of a medical analysis method according to a preferred embodiment of the present invention. FIG. 3 is a block diagram of a medical analysis apparatus according to a preferred embodiment of the present invention. FIG. 3 is a medical analysis method and a medical method according to a preferred embodiment of the present invention. A schematic diagram of the situation in which the device is applied. BRIEF DESCRIPTION OF THE DRAWINGS Figures 1 to 3 illustrate a medical analysis method and an analysis of a preferred embodiment of the present invention. The medical analysis method of the present embodiment includes an image capturing step S, a state determining step S02, a spirometry step, and an analyzing step SG4. The medical analysis device of the present embodiment 1 〇匕 3 shirt image capture module U, - state determination module 12, - breathing volume 2, '13 卩 and 'modeling, address 14. The context system applied in this embodiment The diseased child is subjected to a non-contact respiratory measurement and provides a child condition analysis. Of course, this is merely an example and is not intended to limit the present invention. =, the state determination module 12 and the respiratory measurement module of the present embodiment The group (1) and the module 14 are integrated into a computer system c as an example. First, in the image capturing step, in step S01, the image is captured by the image capturing mode pair - the biological image is captured. u can include a resolution camera to provide a high-resolution image. As can be seen from Figure 3, ^ = no need to contact the f-tester, so the subject can be controlled, and can rest safely. The determining step S02 is determined by the state determining module η according to the images according to the image of 201223505, wherein the living being is in a measurable state or an unmeasurable state. The measurable state may include a completely static or small motion of the creature. For example, small movements of hands and feet do not affect breathing. In the embodiment, the state determining step may include: analyzing a foreground of the image according to the images or - a background; capturing at least the area of the foreground based on the images; and placing the area feature in a judgment model to determine that the creature is located: a measurable state or a non-measurable state. ' In this embodiment, the background is modeled by Gaussian distribution, and the foreground can be separated according to this. The analysis of the foreground or background is based on the pixel of the captured image (pncel), because this part belongs to the prior art. The application, so the details are further described. After analyzing the foreground, at least one area feature of the foreground is captured according to the images. The area features may include the area of the single image and the area of the foreground image of the second image. The amount of change, or the image before and after: product; the amount of change. Among them, the area of the front view of the single-image can reflect the change of the foreground in the continuous image. The area characteristics of the change can also reflect the change in the foreground. ^ The amount of change is increasing, reflecting that the creature may be in motion. For example, the amount of change in the field is getting smaller and smaller, reflecting that the creature may be tending to Static, '=If you turn over the movement, the area characteristics of the mutual exclusion area of the front and back images can also reflect the change of the foreground. When the change of the mutual exclusion area is 201223505, the amount is more than 1 and the image is displayed. The near overlap of the prospects reflects that the bio-fertiles tend to be stationary. Conversely, when the variation of the mutual exclusion area is larger, it reflects that the organism is moving significantly. According to this, the above-mentioned area features can be used to describe the action state of the foreground. After the field area feature is found, the area feature can be placed into a judgment model to determine that the creature is in a measurable state or an unmeasurable state. The judgment model is, for example, an artificial intelligence model such as a Bayesian classifier, a neural network, an adab〇〇St classifier, an SVM classifier, and an LDA linear discriminant analysis (Fisheris-a1, ClaSSifie〇, 0, tsu classification, decision tree). , K-nearest neighbor (KNN) hidden Markov model (hiddenmark〇vm〇dei) and other methods of judgment. By judging that the creature is in a measurable state or the unmeasurable state, the breathing volume can be greatly improved. In addition, in the state judging step, t can determine that the living being is in a transition state between the measurable state and the unmeasurable state. FIG. 4 shows the embodiment. The relationship between the three states used: the measurable state W, the unmeasurable state S2, the transition state S3, and the transition state phantom exists between the unmeasurable state S2 and the measurable state S1, that is, when the creature is substantially The action (unmeasurable state S2) passes through the transition state S3 to the stationary or small towel field action (measurable state S1), or from the transition state S3 to the unmeasurable state S2. By increasing the transition state S3, can The performance of the state judgment is improved, and the correctness of the respiratory measurement is improved. In addition, the invention adds a convergence coefficient to the judgment model, which is changed according to the biologically located state, the unmeasurable state, or the transition state. 201223505 The size is changed to control the convergence speed of the measurable state, the unmeasurable state, or the transition state. The smaller the convergence coefficient, the more sensitive the state judgment is, that is, the easier it is to feel the change, and the convergence speed is slower; the larger the convergence coefficient The less sensitive the state is, the less likely it is to feel the change, and the convergence speed is faster. In this embodiment, the convergence coefficient is the background model parameter, which is used to control the convergence speed of the scene model. For example, as follows: When the system is in the measurable state si (stationary), it can be set that the convergence coefficient is small, so that the system can not be easily changed, which helps to monitor the biological movement; when the system becomes When the state S2 cannot be measured (large action), the r pride coefficient can be set to zero. This can make the system not change due to large movements; In the state S3, the representative organism has slowly stood still. At this time, the convergence coefficient can be adjusted to accelerate the convergence' until the convergence is completed, and the system returns to the measurable state S1. At this time, the number of receipts can be reduced, which helps to continuously monitor the organism. Action. The speed and efficiency of state judgment can be improved by the convergence coefficient. In addition to the above-mentioned state judgment step using the judgment model, the simple judgment method is utilized. In this aspect, the coffee step includes: Image-analyzing the foreground- or background of the image, extracting at least one area feature of the foreground according to the image; and comparing the area characteristic with a threshold value, or the unmeasurable state. The measurement status of the county has application _, for example, as follows, and the area of the single two: the area as the area feature: if S], the area feature (four) will approach the background state of the speculation). When the user starts to move, the face = the image is almost all „ the area feature value will become larger. “The foreground is gradually appearing.” Therefore, the area 7 value will be large (because the product feature right is greater than a certain threshold, the system will 201223505 knows that the user is just moving, and switches to the unmeasurable state s2, and sets the convergence coefficient to zero. When the user moves to rest, the area feature will always maintain a certain value (a is because the foreground has not changed) If the value does not change much for a certain period of time, it means that the user has been stationary, then the system state goes to the transition state S3, and the convergence coefficient value is increased to accelerate the convergence until the acquisition is completed, the system will return The state si ' can be measured and the convergence coefficient value can be lowered. In addition, the present invention can also make (4) he compares the way to know that the living being is in a measurable state or a non-measurable state. For example, in the state judging step, It is not necessary to analyze the foreground or background of the image, but may include: calculating the moving momentum of the biological movement, such as a moving vector (m.-ve coffee); and: judging the using a decision mechanism Whether the moving momentum meets the conditions of the respiratory measurement, that is, the living organism is judged to be in a measurable state or a non-measurable state. Since the motion vector is a prior art, the details thereof are not described here. The decision machine: for example, will move The momentum is compared with the threshold, or the change of the momentum is changed with time, and then the change is compared with the solution and the periodic amplitude. When the state is forwarded, the image can be: According to the method, and the breathing is generated according to the images; the second line is applied to the user. The present invention is applicable to the user lying on the back, lying on the side, standing on the stand, corresponding to these different postures, and the invention of the creatures such as 5 is judged by the first judgment. The breathing direction of the creature, and then generate the call = 12 201223505 suction curve, thereby greatly improving the accuracy of the respiratory measurement information. The following examples illustrate the process of obtaining the breathing direction and the breathing curve. Flow (〇ptical fl〇w ), to obtain the motion vector between the two images, denoted as MU, where X and y represent the coordinates of the element, and j represents the 』th image. And the vertical components are recorded as volumes () and (10) respectively. Then add the following equations respectively ^ and (10) also): ϋ mine ^ | (x, y) ΣΣ - ^ f and generation curve as shown in Figure 5. Where H is the limit value of the y-square of the pixel, and w is the limit value of the u-mark of the element. For the mm as the direction of the breathing, the number of the two variations can be calculated as V^vv. When vhA is at , the direction of the breathing direction, when Vh τ breathing direction. When the action is determined in the vertical direction...& direction, it is recorded as the example of 7乍, and the curve of 矸 is as shown in Fig. 6 =! Suction: line '_ time axis: accumulate (recorded as The action shown): . The (10)7' curve is shown in Figure 7 W, where W is set to zero. &4S;= 13 201223505 With the current point, and as a smooth judgment, where m is the size of the smoothed window, meaning: and take m points forward and backward, a total of 2m + l points, according to. Or, it should be noted that the above-mentioned breathing direction is divided into horizontal and vertical directions, and only the present invention is recorded. The present invention is different in meaning and different in breathing direction due to different conditions. After the breathing curve is obtained, the analysis step sq4 can be further performed by analyzing the biological condition according to the respiratory measurement result by the analysis module, and the analysis can include calculating the respiratory characteristics, statistics, making a report or a condition. Probability prediction, or diagnosis. For example, the bioanalysis is sleep apnea, dying of dying patients, asthma, apnea, transitional ventilation syndrome, adult respiratory distress, pulmonary embolism, acute pulmonary edema, lung arrest, or morphine poisoning and others. Respiratory related symptoms. The above symptoms are manifested in the following conditions: the initial judgment of the sleep breathing industry (the chest and the abdomen breath will be inconsistent, and there will be characteristics of respiratory arrest); the death of the patient is determined by the end of the breath (the depth of breathing and the breath of both breaths) Abnormal time interval; asthma (dyspnea); sudden apnea and sudden death (breathing stop); transient ventilation syndrome (breathing too fast, too deep); adult respiratory distress (breathing deep and fast); pulmonary embolism (breathing aggravation) , frequency is accelerated); heart sputum pulmonary edema (short breath); cardiopulmonary arrest (decreased breathing rate) morphine poisoning (reduced breathing rate). The invention can obtain data by breathing curve, which includes at least respiratory intensity (amplitude), frequency, time, trend, etc., or other auxiliary materials to obtain accurate judgment. In summary, the medical analysis method and medical analysis device of the present invention are 14
I 201223505 採用非接觸式的呼吸量測方式，藉由對生物進行影像擷 取，並藉由分析影像而得到該生物之呼吸狀況，進而利用 呼吸狀況刀析。亥生物狀況。本套人性化智慧方法乃使用非 接觸式之視訊影像分析技術，用以量測使用者的呼吸資 訊。使用者不需改變任何生活習慣與擔心儀器影響活動自 由’系統即會自動分析使用者的呼吸情境，進而動態與精 準的量測使用者之呼吸資訊。此外，本發明在呼吸量測之 前，會先判斷該生物位於可量測狀態（例如靜止或小幅動 •作）、.或是不可量測狀態（例如大幅動作），以致所得到的 呼吸量測資訊之可靠度及精確度皆大幅提升，致使本發明 可應用更廣泛及精確的病症分析。此方法及裝置不僅適合 於居家和遠距照護，更可以提供安養機構和醫療中心做呼 吸資訊量測，以做病症和健康的輔助判斷與監控。 以上所述僅為舉例性，而非為限制性者。任何未脫離 本發明之精神與範嘴，而對其進行之等效修改或變更，均 應包含於後附之申請專利範圍中。 【圖式簡單說明】 圖； 圖1為本發明較佳實施例之一種醫學分析方法的流程 圖2為本發日緣佳實關之—種^ 示意圖； 圖3為本發明較佳實施例之醫學分析方法及醫學分析 裝置應用的情境示意圖； 15 201223505 圖4係顯示本實施例所使用之三種狀態的關係；以及 圖5至圖7為呼吸量測的曲線圖。 【主要元件符號說明】 10 :醫學分析裝置 11 :影像擷取模組 12:狀態判斷模組 13 :呼吸量測模組 14 :分析模組 C :電腦系統 S01〜S04:醫學分析方法步驟 16I 201223505 uses a non-contact spirometry method to capture the organism's respiratory condition by analyzing the image and then using the respiratory condition. Hai biological situation. This set of humanized wisdom methods uses non-contact video image analysis technology to measure the user's breathing information. The user does not need to change any living habits and worry about the instrument's influence on the freedom of the activity. The system automatically analyzes the user's breathing situation, and then dynamically and accurately measures the user's breathing information. In addition, the present invention determines whether the living being is in a measurable state (eg, static or small motion), or an unmeasurable state (eg, a large motion) before the spirometry, so that the obtained respiratory measurement is obtained. The reliability and accuracy of the information are greatly enhanced, so that the present invention can be applied to a broader and more accurate analysis of the condition. The method and device are not only suitable for home and remote care, but also provide ancillary institutions and medical centers for respiratory information measurement for auxiliary judgment and monitoring of illness and health. The above is intended to be illustrative only and not limiting. Any equivalent modifications or alterations of the present invention are intended to be included in the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flow chart of a medical analysis method according to a preferred embodiment of the present invention; FIG. 3 is a schematic diagram of a Japanese syllabus; FIG. 3 is a schematic view of a preferred embodiment of the present invention; A schematic diagram of a medical analysis method and a medical analysis device application; 15 201223505 Figure 4 shows the relationship of the three states used in the present embodiment; and Figures 5 to 7 are graphs of the respiratory measurement. [Major component symbol description] 10: Medical analysis device 11: Image capture module 12: State judgment module 13: Respiratory measurement module 14: Analysis module C: Computer system S01~S04: Medical analysis method step 16