201102941 ' 六、發明說明： 【發明所屬之技術領域】 本發明係有關於一種照護裝置，更詳而言之，係關於 一種可對各種生理訊號進行筛選、分析與訓練以降低資料 複雜度並產生對應之特徵值之照護裝置。 【先前技術】 隨著醫療科技的進步，人類的平均壽命已逐漸延長， 使社會已逐漸步入高齡化結構，相對地，受到目前生育率 *普遍降低的少子女化趨勢影響，造成越來越多的高齡者無 法藉由家人與家庭獲得妥善充分的照顧。另一方面，現代 人常因面臨工作競爭的心理壓力，輕忽對自身健康管理， 例如，飲食與生活起居的不規律，甚至因延誤就醫以致病 情惡化難治的現象，使積勞成疾與慢性病患有年輕化的趨 勢，因此，在照護需求增加而照護資源卻相當有限（例如： 照護人力負荷過大或床位不足）的狀況下，常造成醫療品 Φ質無法兼顧。 對此，為了在有限的醫療人力與資源條件下，即時掌 握每個老年人或慢性病患的健康資訊，以提昇醫療品質、 照護便利性與舒適性，並舒缓家屬或專業醫療人員之照護 壓力等目的，近年來已有諸多學者或業者結合感測與通訊 技術進行遠距照護系統的相關研究，如我國專利公告號第 1255703號「遠端照護系統」一案，其主要係揭示一種可 將藍芽血壓計所量測的資訊，以主動或被動的方式透過網 際網路傳送至遠端伺服器的技術，然，此遠端照護系統僅 111320 201102941 量測有血堡#料’且其所量測的資料係直接透過本地資料 庫，一储存，再傳送至遠端贿器，而未經由初步過滤筛 l私序目此除了需要耗費遠端飼服器的資源進一步分析 外’若儲存的資料量過大將會造成更新傳送緩慢或因資料 庫的館存容量不夠等資料處理上的_。 加又如我國專利公告號第M296686號「遠距醫療系統 木構」t ’其主要係揭示-種將所量測的生理數值透過 數位機上盒傳送至遠端資訊系統平台的技術，_，此系統 雖然可同時量測不_生理訊號，但由於此生理訊號未經 由初步過渡篩選程序，且受限於頻寬限制而只能傳送 生理訊息，使得遠端資㈣統平台仍需要耗f額外資二進 行分析處理，並且喪失處理先機。 此外’為了縮減傳輸育料量，雖可免丨田次" 守干』κ rr里雏j利用賁料壓縮（扣忪 compression)技術來減少資料儲存的佔用空間或頻寬，然 而，經壓縮後的生理資料於照護管理端進行資料還^時二 仍會因部分資料的遺失而導致些微的訊號失真，特別是對 於攸關生命的指標生理訊號（例如：心率變異度、心房/ 心室顫動），因此基於精確度與安全性的考量，將不宜直接 進行資料縮減’且還原後的資料仍需佔用照護管理端的處 理資源。 地 職是，如何提供一種可對各種生理訊號之進行篩選' 分析與訓練之照護裝置’能有效降低資料複雜度，以改盖 習知照護系統的資料傳送速度緩慢、資料儲存容量不夠^ 處理資源耗費等情況，實為目前此生 醫產業應用中亟待解 Π1320 4 201102941 決之問題。 【發明内容】 鑒於上述習知技術之缺點，本發明係提供一種照護裝 置，係連結至少一照護平台，該照護裝置係包括：訊號擷 取模組，係感測人體的生理反應或環境狀態以產生訊號資 料；處理模組，係具有運算判斷規則單元，用以透過該運 算判斷規則單元將該訊號擷取模組所產生之該訊號資料進 行篩選與分析以降低該訊號資料之複雜度，並持續透過該 •運算判斷規則單元訓練及產生對應該訊號資料之特徵資 訊；以及通訊模組，用以將該處理模組所產生之該特徵資 訊傳送至該照護平台，俾使該照護平台依據該特徵資訊執 行相應之照護工作。 於一較佳實施例中，本發明所提供之照護裝置係連接 至少一輔助設備，其中，該輔助設備可為電動輪椅、助行 器、電動床或其他醫療輔助設備。而所述之照護裝置復包 鲁括一控制模組，係用以透過命令訊號控制該輔助設備，其 中’該照護平台可依據所接收之該特徵資訊將該命令訊號 傳送至該控制模組，以相應地控制該輔助設備益或是相應 地控制鄰近該照護裝置之環境設備，例如燈具、收音機、 電視、風扇、空調機、警報器或其他非醫療設備。 於另一較佳實施例中，上述之運算判斷規則單元可透 過模糊控制演算法（Fuzzy control Algorithm )、類神經網 路茂异法（Neural Network Algorithm )或小波轉換演算法 (Wavelet Transform Algorithm)進行該訊號資料之篩選、分 5 Π1320 201102941 ^ 析或學習以產生相應之該特徵資訊。詳言之，該類神經網 路演算法為可調控式部分最小平方演算法（Partial Regularized Least Squares Algorithm) ° 相較於習知技術，本發明所提供之照護裝置可將所感 測之人體生理訊號，透過運算判斷規則單元進行篩選、分 析與訓練以產生相應之特徵資訊。藉此，可有效降低資料 複雜度並縮減資料量，以改善習知技術中資料傳送頻寬不 足、上傳速度缓慢、儲存容量不足及耗費大量處理資源的 情況。 【實施方式】 以下係藉由特定的具體實例說明本發明之實施方 式，熟悉此技藝之人士可由本說明書所揭示之内容輕易地 暸解本發明之其他優點與功效。本發明亦可藉由其他不同 的具體實例加以施行或應用，本說明書中的各項細節亦可 基於不同觀點與應用，在不悖離本發明之精神下進行各種 修飾與變更。 以下之實施例係進一步詳細說明本發明之觀點，但並 非以任何觀點限制本發明之範疇，以下圖式僅以簡化之示 意圖式說明本發明之基本構想，遂圖式中僅例示與本發明 有關之元件而非按照實際實施時之元件數目、形狀及尺寸 繪製，因此再實際實施時，各元件之型態、數量及比例並 非以圖式為限，可依實際設計需要作變化，合先敘明。 如第1圖所示，係顯示本發明之照護裝置之架構圖。 如圖所示，本發明之照護裝置1係連結至少一照護平台2。 6 Π1320 201102941 具體而言，該照護平台2可為具有通訊功能之電腦或伺服 器，並可連結醫院、急診中心、養護中心或其他醫療照護 介面。照護裝置1包括：訊號擷取模組11、處理模組 以及通訊模組13。以下對本發明所揭之照護裝置1之各組 成構件進行詳細說明。 該訊號擷取模組11，可為影音偵測器、生理訊號偵測 器及/或環境訊號偵測器，係用以感測各種人體的生理反應 鲁或環境狀態並藉以產生訊號資料。以生理訊號偵測器為 例’該訊號資料可為心電圖、血壓、血氧濃度'血糖濃度、 體溫值或其他生理訊號。例如，心電圖量測裝置可作為— 種生理訊號偵測器，係將電極片夾置於人體四肢或胸前， 以透過四肢及胸前皮膚表面取得心肌活動所散發出之電性 >舌動的不同頻率、間隔的相應變化，俾產生規律與非規律 的類比訊號。 該處理模組12，係具有運算判斷規則單元121，用以 籲依據該運算判斷規則單元ι21將該訊號擷取模組η所產生 之該訊號資料進行筛選與分析以降低該訊號資料之複雜 度’並持續透過該運算判斷規則單元121訓練及產生相應 之特徵資訊。具體而言，處理模組12可為中央處理器或可 程式竣輯控制器（Programmable Logic Controller)，用以 將不同人體生理反應所產生的多個訊號資料進行篩選與分 析’並持續進行訓練與學習，藉以找出最佳化的特徵資訊。 該通訊模組13，用以將該處理模組12所產生之該特 徵貢訊傳送至該照護平台2，俾使該照護平台2依據該特 7 111320 201102941 徵資訊執行相應之照護工作，其中，通訊模組13可透過有 線方式或無線方式進行資料傳輸。具體而言，常見的有線 傳輸方式可例如為IEEE 1394、USB等有線傳輸介面，而 無線傳輸方式則可例如為區域網路系統（Wireless Local Area Network, WLAN)、蜂巢式行動通訊系統（Global System for Mobile Communication，GSM)、整合封包無線服 務系統(General Packet Radio Service，GPRS)、全球互通微 波存取技術（Worldwide Interoperability for Microwave Access, WiMax)網路系統、個人通訊系統（personai Handyphone System，PHS)、第三代行動通訊技術（3rd Generation，3G)、IEEE 802.U、ZigBee 系統等通訊協定等， 但不以此為限。 於一較佳態樣中，上述之照護平台2依據該特徵資訊 可相應執行被照護者的生理狀態監視、被照護者周圍之環 境狀態監視、被照護者之生理資料分析及比對、遠端控制 被照護者周圍之設備或於緊急狀況發生時發布告警訊•並 通知醫療中心派員處理。 在具體實施上’該運算判斷規則單元121可透過模糊 控制演算法（Fuzzy control Algorithm)、類神趣網路廣曾法 (Neural Network Algorithm)或小波轉換演算法^ t201102941 ' VI. Description of the invention: [Technical field to which the invention pertains] The present invention relates to a care device, and more particularly to a screening, analysis and training of various physiological signals to reduce data complexity and A care device that produces a corresponding feature value. [Prior Art] With the advancement of medical technology, the average life expectancy of human beings has gradually prolonged, and society has gradually entered an aging structure. Relatively, it is affected by the trend of fewer children who are currently generally reduced in fertility rate, resulting in more and more Many elderly people cannot get proper and adequate care from their families and families. On the other hand, modern people often face the psychological pressure of work competition, and neglect their own health management, for example, the irregularity of diet and daily life, and even the delay in medical treatment, which makes the disease worse and difficult to treat, making overwork and chronic diseases. There is a trend of rejuvenation. Therefore, when the demand for care increases and the care resources are quite limited (for example, the care load is too heavy or the bed is insufficient), the medical products are often unable to balance. In this regard, in order to control the health information of each elderly or chronic patients under limited medical manpower and resources, to improve medical quality, care convenience and comfort, and to relieve the care pressure of family members or professional medical personnel. For other purposes, in recent years, many scholars or practitioners have combined sensing and communication technologies to conduct related research on remote care systems. For example, in the case of "Remote Care System" in China Patent Publication No. 1255703, it mainly reveals a kind of The information measured by the Bluetooth sphygmomanometer is transmitted to the remote server through the Internet in an active or passive manner. However, this remote care system only measures 111320 201102941 and has a blood castle #料' The measured data is directly transmitted through the local database, and then stored to the remote bribe, without passing through the preliminary filter screen. In addition to the need to further analyze the resources of the remote feeder, the data is stored. If the amount of data is too large, the update will be slow or the data will be processed due to insufficient library capacity. For example, China Patent Publication No. M296686 "Television Medical System Wood Construction" t' is mainly to reveal the technology of transmitting the measured physiological values to the remote information system platform through the digital machine box, _, Although this system can measure non-physiological signals at the same time, since this physiological signal has not passed the initial transition screening procedure and is limited by the bandwidth limitation, only the physiological information can be transmitted, so that the remote (4) platform still needs to consume a large amount of money. Foreign capital II conducts analysis and processing, and loses the opportunity to deal with it. In addition, in order to reduce the amount of transmission and breeding, it is possible to reduce the space or bandwidth of data storage by using the compression compression technology. After the physiological data is carried out on the care management side, the data will still be slightly distorted due to the loss of some data, especially for the physiological signals of vital signs (eg heart rate variability, atrial/ventricular fibrillation). Therefore, based on the accuracy and security considerations, it is not appropriate to directly reduce the data' and the restored data still needs to occupy the processing resources of the care management terminal. The job is to provide a kind of care device that can filter various physiological signals. The analysis and training of the device can effectively reduce the complexity of the data, so that the data transfer speed of the conventional care system is slow and the data storage capacity is insufficient. The cost and other circumstances are indeed the problems that need to be solved in the application of this biomedical industry at 1320 4 201102941. SUMMARY OF THE INVENTION In view of the above disadvantages of the prior art, the present invention provides a care device that is coupled to at least one care platform, the care device comprising: a signal capture module that senses a physiological response or an environmental state of the human body. Generating a signal data; the processing module has a computing judgment rule unit for filtering and analyzing the signal data generated by the signal capturing module by the computing unit to reduce the complexity of the signal data, and And continuously, through the operation, the rule unit is trained to generate feature information corresponding to the signal data; and the communication module is configured to transmit the feature information generated by the processing module to the care platform, so that the care platform is based on the Feature information performs the corresponding care work. In a preferred embodiment, the care device of the present invention is coupled to at least one auxiliary device, wherein the auxiliary device can be an electric wheelchair, a walker, an electric bed, or other medical aid. And the protection device is configured to control the auxiliary device by using a command signal, wherein the service platform can transmit the command signal to the control module according to the received feature information. The auxiliary equipment is controlled accordingly or the environmental equipment adjacent to the care device, such as a light fixture, a radio, a television, a fan, an air conditioner, an alarm or other non-medical equipment, is controlled accordingly. In another preferred embodiment, the operation determining rule unit may be performed by a fuzzy control algorithm, a neural network algorithm, or a Wavelet Transform Algorithm. The screening of the signal data is divided into 5 Π 1320 201102941 ^ or learned to generate the corresponding information of the feature. In detail, the neural network algorithm is a Partial Regularized Least Squares Algorithm. Compared with the prior art, the care device provided by the present invention can sense the physiological signal of the human body. Screening, analysis and training are performed through the operation judgment rule unit to generate corresponding feature information. This can effectively reduce the complexity of the data and reduce the amount of data, so as to improve the data transmission bandwidth, slow upload speed, insufficient storage capacity and a large amount of processing resources in the prior art. [Embodiment] The embodiments of the present invention are described below by way of specific examples, and those skilled in the art can readily understand other advantages and effects of the present invention from the disclosure of the present disclosure. The present invention may be embodied or applied by other specific embodiments, and various modifications and changes may be made without departing from the spirit and scope of the invention. The following examples are intended to describe the present invention in further detail, but are not intended to limit the scope of the present invention. The following drawings illustrate the basic concept of the invention only in a simplified schematic diagram, which is merely illustrative of the invention. The components are not drawn according to the number, shape and size of the components in actual implementation. Therefore, in actual implementation, the types, quantities and proportions of the components are not limited to the drawings, and can be changed according to the actual design needs. Bright. As shown in Fig. 1, a schematic view of the care device of the present invention is shown. As shown, the care device 1 of the present invention is coupled to at least one care platform 2. 6 Π1320 201102941 Specifically, the care platform 2 can be a computer or server with communication functions and can be connected to a hospital, emergency center, maintenance center or other medical care interface. The care device 1 includes a signal capture module 11, a processing module, and a communication module 13. Hereinafter, each component of the care device 1 disclosed in the present invention will be described in detail. The signal capture module 11 can be a video detector, a physiological signal detector and/or an environmental signal detector for sensing physiological responses of various human bodies or environmental conditions and generating signal data. Take a physiological signal detector as an example. The signal data can be an electrocardiogram, blood pressure, blood oxygen concentration, blood glucose concentration, body temperature value or other physiological signals. For example, the electrocardiograph can be used as a physiological signal detector to place the electrode sheets on the limbs or chest of the human body to obtain the electrical properties emitted by the myocardial activity through the limbs and the skin surface of the chest. Corresponding changes in different frequencies and intervals, 俾 generate regular and irregular analog signals. The processing module 12 has an operation determining rule unit 121 for requesting the filtering and analysis of the signal data generated by the signal capturing module η according to the operation determining rule unit ι21 to reduce the complexity of the signal data. And continuously, through the operation judgment rule unit 121, training and generating corresponding feature information. Specifically, the processing module 12 can be a central processing unit or a programmable logic controller (Programmable Logic Controller) for filtering and analyzing multiple signal data generated by different human physiological responses' and continuously training and Learn to find out the best feature information. The communication module 13 is configured to transmit the feature information generated by the processing module 12 to the care platform 2, so that the care platform 2 performs corresponding care work according to the information of the special 7111320 201102941, wherein The communication module 13 can transmit data by wire or wirelessly. Specifically, a common wired transmission method may be, for example, a wired transmission interface such as IEEE 1394 or USB, and the wireless transmission method may be, for example, a Wireless Local Area Network (WLAN) or a cellular mobile communication system (Global System). For Mobile Communication, GSM), General Packet Radio Service (GPRS), Worldwide Interoperability for Microwave Access (WiMax) network system, personal communication system (personai Handyphone System, PHS) Third-generation mobile communication technology (3rd Generation, 3G), IEEE 802.U, ZigBee system and other communication protocols, but not limited to this. In a preferred aspect, the above-mentioned care platform 2 can perform physiological state monitoring of the caregiver, environmental state monitoring around the caregiver, physiological data analysis and comparison of the caregiver according to the characteristic information, and the distal end. Control equipment around the caregiver or issue an alert when an emergency occurs. • Inform the medical center to dispatch a member. In the specific implementation, the operation judging rule unit 121 can transmit a fuzzy control algorithm, a neural network algorithm, or a wavelet transform algorithm.
Transform Algorithm)等智慧型演算法對訊號栂取模组^ 所取樣的類比訊號資料進行篩選，當訊號資料或力I丨練 (training)樣本越多、越正確、差異性越大，智犛型演蓄去 的能力就越強，以於訊號資料不失真的前提下，筛選與八 111320 201102941 析出對應的特徵資訊，並持續進行訓練與學習以保持該特 徵資訊的最佳化，俾達到有效縮減資料量與降低資料複雜 度的目的。 於另一具體實施例中，本發明之照護裝置復包括一控 制模組，可例如為遙控器或控制按鍵，且該照護裝置係由 該通訊模組連結至少一輔助設備或環境設備，用以將該控 制訊號透過該通訊模組傳送至該輔助設備或環境設備，且 該補助設備或環境設備依據所接收之該控制訊號執行對應 ®之控制功能命令。具體而言，該輔助設備可為電動輪椅、 電動床或其他醫療輔助設備。該環境設備可為電燈、風扇、 收音機、電視、空調、揚聲器或其他非醫療設備之其中至 少一者。 舉例而言，以輔助設備是電動輪椅，環境設備是空調 為例，其電動輪椅控制功能命令可例如為設備開啟、設備 關閉、方向調整、速度調整、進退動作、轉彎動作等。空 $調控制功能命令可例如為設備開啟、設備關閉或風量數值 調控等，但不以此為限，且該輔助設備或環境設備之控制 狀態透過該通訊模組傳送至該照護平台。 於再一實施態樣中，可於電動輪椅中設置一坐墊壓力 感測器，該坐墊壓力感測器可持續偵測使用者的因坐姿所 產生之壓力訊息，透過本發明之運算判斷規則單元對該壓 力訊息進行篩選並擷取出對應之特徵值，再以控制模組設 定該電動輪椅之坐墊壓力感測器的荷重分布位置或控制一 警報單元發出聲響以提醒使用者改變坐姿。 9 111320 201102941 另外，前述之照護裝置亦可包括影音擷取模組，用以 擷取行為特性資料，並依據該運算判斷規則將該行為特性 進行篩選與分析以產生相應之特徵資訊，其中該影音擷取 模組可為攝影機、視訊會議攝影機、麥克風等，而該行為 特性可為人臉、四肢動作或聲音等。舉例而言，以影音擷 取模組是攝影機為例，係可透過擷取人臉的表情或聲音的 特徵資訊傳送至照護平台，俾利於照護人員掌握非自然生 理狀態的病患需求，且經由影音擷取模組所擷取的行為特 性資料同樣藉由處理模組依據運算判斷規則進行篩選與分 析，俾達到有效縮減資料量之效果。 請參閱第2圖，係本發明之照護裝置之一具體實施例 的示意圖。如圖所示，被照護者坐在電動輪椅31上，不斷 透過影音偵測器310取得被照護者的臉部特徵影像、四肢 動作影像或各種生理活動所產生之聲音。另外，利用生理 訊號偵測器311取得被照護者的各種生理訊號如心電圖、 血壓、血氧濃度、血糖濃度或體溫值。接著，將影音訊號 及生理訊號透過處理模組312進行筛選與分析以降低該訊 號資料之複雜度，並持續進行訓練與學習，藉以找出最佳 化的特徵資訊。 以上述步驟降低資料量及資料複雜度後，再透過通訊 模組313利用無線網路4將特徵資訊傳至醫院照護監控中 心5。因此，醫院照護監控中心5可即時監看被照護者的 狀態。於較佳態樣中，本發明之照護裝置復包括一控制模 組，係用以透過命令訊號控制圖示中的電動輪椅31，而醫 201102941 院照護監控中心5可依據所接收之該特徵資訊將該命令訊 號傳送至該控制模組以相應地控制該電動輪椅31，亦或是 相應地控制鄰近該電動輪椅31之環境設備，如為燈具32 或警報器33。本發明可將多種資訊整合在一起進行篩選與 分析，例如一維度(ECG、Sp02等）噪音消除或二維度（臉部） 特徵值判斷與降低資訊複雜度，以將所產生的特徵資訊即 時傳送給多個遠方照護者，達到多方即時關懷等目的。此 外，透過雙向訊息的傳送，除了將資訊傳遞給照護者外， 也能動態遠端操控照護環境，進一步提供被照護者生活上 之協助。 以下透過另一具體實施例說明本發明之照護裝置的 運作方式。以ECG訊號的偵測為例，訊號擷取模組可透過 一心電圖偵測器來實現，藉以獲得ECG訊號。接著，將 ECG訊號輸入處理模組進行篩選與分析以降低該ECG訊 號資料之複雜度，並產生對應之特徵資訊。然後，再透過 通訊模組，用以將該處理模組所產生之該特徵資訊傳送至 該照護平台。最後，透過儲存於該照護平台中的人工智慧 系統即時判斷ECG訊號，並與個人過去健康紀錄比較，以 產出生理综合指標與心臟變異警訊。若一但發生異常情 況，則透過照護平台遠端控制被照護者所在位置之警報 器，使醫護人員能立即採取行動。 請參閱第3圖，係顯示本發明以類神經網路技術的調 控式部分最小平方演算法（Partial Regularized Least Squares Algorithm, PRLS)之架構示意圖。如圖所示，係顯 】] 111320 201102941 示三層式類神經網路之實施架構圖，該處理模組係預設有 可調控參數λ，且該可調控式部分最小平方演算法結構係 可包括·輸入層、1½藏層以及輸出層，其中，該輸入斤传 依據多個該訊號資料而產生對應的自變數集合（Χι，χ2… Xm} ’其中Xi= [x"，h/，…，xm]T ’ η表示運算的樣本數，m 表示輸入層的變數大小；該隱藏層係將先自變數集入& &，…，xm}依據部分最小平方分析法(partialLeast Squares， PLS)分別產生對應的權重值以，…，户以丁， ……’ …，Ρη^τ’ k g示隱藏層的 變數大小，並對各該對應權重值進行線性映射運算求得隱 藏層數值，然後，再依據最小平方法(Least Square，Ls)與 該可調控參數又調整而產生對應的權重值 至最佳化，並加總對應至應變數{>?1，力，“·，、}，以產生"相 應之特徵資訊。 更具體而言，上述之可調控參數λ可為利用貝氏定理 (Bale’s rule )中所求得的後驗機率（p〇sted〇r ㈣ 值，藉以修正慣用的前驗機率（pri〇r pr〇babili^)所產生的 主觀判斷上的誤差，並進而利用到過濾多餘的資料，俾達 到有效縮減資料量之效果。 在具體實施上，該特徵資訊可為有限數值、特定波 形、頻率、振幅的變異值，利用調控式部分最小平分演算 去對Λ號資料進行篩選與分析程序後，所得出的特徵資訊 :大里降低貝料複雜度。舉例而言，上述之運算判斷規則 單元可預設僅操取訊號時域區間的變異值，以連續量測心 1]]320 201102941 跳次數三分鐘的分析波形為 分鐘㈣下、第3分鐘跳58下，：知跳50下、第2 Γ〇 . 下則處理模組僅擷取變異 值[、8]的相關波形進行篩選盥分拚，廿41：入^ 波形資料，以有效縮減資料量，更新二：非王部㈣兆 不夠、額外處理㈣歸❾U新傳魏慢、儲存容量 邻八4圖，係顯示本發明之照護裝置利用調控式 平方演算法進行特徵值擷取之流程圖。如圖所 不’百先’於步驟S40，利用苑执 > 朴 預叹之特徵值執行對輸入變 數初七化的動作。於步驟S4卜透過部分最小平方法計算 輸入义數之核重。接著，於步驟S42，透過線性映射食可 調式部分最小平方法計算特徵值之權重。於步驟如，利 用該特徵值之權重更_輸人變數與料難，並域認是 否符合限制條件，若不符合，則返回步驟⑽，若符人， 則進至步驟S44,停止計算並輸出該更新之特徵值。口 综上所述’本發明所提供之照魏置相較於習知技 術’至少包括下述功效： ⑴可有效地縮減非必要的資料量，俾解決健存容 量不夠的問題。 (2)可降低資訊的複雜度’以避免更新傳送速度緩 慢以及處理資源的耗費。 上述實施例僅例示性說明本發明之原理及其功效，而 非用於限制本發明。任何熟習此項技藝之人士均可在不違 背本發明之精神及範脅下，對上述實施例進行修飾與改 變。因此，本發明之權利保護範圍，應如後述之申請專利 Π1320 13 201102941 範圍所列。 【圖式簡單說明】 第1圖係本發明之照護裝置之架構圖； 第2圖係本發明之照護裝置之一具體實施例的示意圖； 第3圖係顯示本發明以類神經網路技術的調控式部分 最小平分演算法之架構示意圖；以及 第4圖係顯示本發明之照護裝置利用調控式部分最小 平分演算法進行特徵值擷取之流程圖。 【主要元件符號說明】 1 照護裝置 2 照護平台 11 訊號擷取模組 12、 312 處理模組 121 運算判斷規則單元 13、 313 通訊模組 3 照護病房 31 電動輪椅 310 影音偵測器 311 生理訊號偵測器 32 燈具 33 警報器 4 無線網路 S40〜S44 5 醫院照護監控中心 步驟 14 Π1320Transform Algorithm) and other intelligent algorithms filter the analog signal data sampled by the signal acquisition module ^. When the signal data or force I trains, the more samples, the more correct, the greater the difference, the wisdom type The stronger the ability to perform, the better the information is not distorted, the feature information corresponding to the analysis of the eight 111320 201102941, and the continuous training and learning to maintain the optimization of the feature information, to achieve effective Reduce the amount of data and reduce the complexity of the data. In another embodiment, the device of the present invention includes a control module, which may be, for example, a remote controller or a control button, and the device is connected to the at least one auxiliary device or the environment device by the communication module. The control signal is transmitted to the auxiliary device or the environmental device through the communication module, and the auxiliary device or the environmental device executes the control function command corresponding to the control signal according to the received control signal. In particular, the auxiliary device can be an electric wheelchair, an electric bed or other medical aid. The environmental device can be at least one of a light, a fan, a radio, a television, an air conditioner, a speaker, or other non-medical device. For example, the auxiliary device is an electric wheelchair, and the environmental device is an air conditioner. The electric wheelchair control function command can be, for example, device opening, device closing, direction adjustment, speed adjustment, advancing and retracting action, turning action, and the like. The air control function command can be, for example, a device-on, a device-off, or an air volume control, but is not limited thereto, and the control state of the auxiliary device or the environment device is transmitted to the care platform through the communication module. In another embodiment, a seat cushion pressure sensor can be disposed in the electric wheelchair, and the seat cushion pressure sensor can continuously detect the pressure information generated by the user in the sitting posture, and the calculation rule unit is operated by the present invention. The pressure message is filtered and the corresponding feature value is extracted, and the control module sets the load distribution position of the seat cushion pressure sensor of the electric wheelchair or controls an alarm unit to sound to remind the user to change the sitting posture. 9 111320 201102941 In addition, the foregoing care device may further comprise an audio and video capture module for extracting behavior characteristic data, and filtering and analyzing the behavior characteristic according to the operation determination rule to generate corresponding feature information, wherein the audio and video The capture module can be a camera, a video conferencing camera, a microphone, etc., and the behavior can be a face, a limb movement or a sound. For example, the video capture module is a camera, which can be transmitted to the care platform by capturing the facial expression or the characteristic information of the voice, thereby facilitating the caregiver to grasp the patient's needs in the unnatural physiological state, and The behavioral characteristics data captured by the AV capture module is also filtered and analyzed by the processing module according to the operation judgment rule, so as to achieve the effect of effectively reducing the amount of data. Referring to Figure 2, there is shown a schematic view of one embodiment of the care device of the present invention. As shown in the figure, the care receiver sits on the electric wheelchair 31 and continuously obtains the facial feature image, the limb motion image or various physiological activities generated by the caregiver through the video detector 310. Further, the physiological signal detector 311 is used to obtain various physiological signals of the care recipient such as an electrocardiogram, blood pressure, blood oxygen concentration, blood sugar concentration or body temperature value. Then, the video signal and the physiological signal are filtered and analyzed by the processing module 312 to reduce the complexity of the signal data, and the training and learning are continuously performed to find the optimized feature information. After the above steps are used to reduce the amount of data and the complexity of the data, the feature information is transmitted to the hospital care monitoring center 5 via the communication module 313 using the wireless network 4. Therefore, the hospital care monitoring center 5 can instantly monitor the status of the care recipient. In a preferred embodiment, the care device of the present invention further includes a control module for controlling the electric wheelchair 31 in the figure through the command signal, and the hospital 201102941 hospital care monitoring center 5 can receive the characteristic information according to the received information. The command signal is transmitted to the control module to control the electric wheelchair 31 accordingly, or to control the environmental equipment adjacent to the electric wheelchair 31, such as the luminaire 32 or the alarm 33. The invention can integrate various information for screening and analysis, such as one-dimensional (ECG, Sp02, etc.) noise elimination or two-dimensional (face) feature value judgment and reduce information complexity, so as to transmit the generated feature information instantaneously. Give multiple remote caregivers to achieve multi-party immediate care and other purposes. In addition, through the transmission of two-way messages, in addition to transmitting information to caregivers, the remote care environment can be dynamically controlled to further provide assistance to the caregivers. The mode of operation of the care device of the present invention will now be described by way of another specific embodiment. Taking ECG signal detection as an example, the signal acquisition module can be implemented by an ECG detector to obtain an ECG signal. Then, the ECG signal is input into the processing module for screening and analysis to reduce the complexity of the ECG signal data and generate corresponding feature information. Then, the communication module is further configured to transmit the feature information generated by the processing module to the care platform. Finally, the ECG signal is instantly determined by the artificial intelligence system stored in the care platform and compared with the individual's past health record to produce a physiologically integrated indicator and a cardiac variability alert. If an abnormal situation occurs, the alarm is placed at the remote end of the care platform to enable the medical staff to take immediate action. Referring to Fig. 3, there is shown a schematic diagram of the architecture of the Partial Regularized Least Squares Algorithm (PRLS) of the neural network technology of the present invention. As shown in the figure, the system shows]] 111320 201102941 shows the implementation architecture diagram of the three-layer neural network. The processing module is pre-configured with a controllable parameter λ, and the configurable partial least squares algorithm structure can be The input layer, the 11⁄2 hidden layer and the output layer are included, wherein the input is generated according to a plurality of the signal data to generate a corresponding set of independent variables (Χι, χ2... Xm} 'where Xi=[x",h/,... , xm]T ' η represents the number of samples of the operation, m represents the variable size of the input layer; the hidden layer is the first self-variable set into &&,..., xm} according to partial least squares analysis (partialLeast Squares, PLS ) respectively generating the corresponding weight value to, ..., the household to D, ...... ', Ρ η ^ τ ' kg shows the variable size of the hidden layer, and linearly map each of the corresponding weight values to obtain the hidden layer value, and then Then, according to the least square method (Least Square, Ls) and the adjustable parameter, the corresponding weight value is optimized to be optimized, and the total is corresponding to the strain number {>?1, force, "·,,} To generate "corresponding feature information. More Specifically, the above-mentioned controllable parameter λ may be a posterior probability (p〇sted〇r (four) value obtained by using Bale's rule, thereby correcting the conventional a priori probability (pri〇r pr〇) Babili^) The subjective judgment error, and then used to filter the excess data, to achieve the effect of effectively reducing the amount of data. In specific implementation, the feature information can be finite value, specific waveform, frequency, amplitude The variability value, using the least partial calculus of the regulation part to filter and analyze the nickname data, the characteristic information obtained: Dali reduces the complexity of the material. For example, the above operation judgment rule unit can be preset only The variability of the time-domain interval of the signal is measured by the continuous amount of 1]] 320 201102941 The analysis waveform of the three-minute hop count is minutes (four), and the third minute is 58 hops: 50 hops, 2 Γ〇. The processing module only extracts the relevant waveforms of the variogram [, 8] for screening and splitting, 廿41: input ^ waveform data to effectively reduce the amount of data, update two: non-king (four) mega is not enough, additional processing (four) Return U Xinwei Wei slow, storage capacity adjacent to 8 4, which shows the flow chart of the illuminating device of the present invention using the controlled square algorithm to perform the feature value extraction. As shown in the figure, the '100 first' is used in step S40. > The characteristic value of Park Pre-sighs performs the action of initializing the input variable. In step S4, the kernel weight of the input semantic number is calculated by the partial least-square method. Then, in step S42, the linear mapping of the food-adjustable part is minimized. The method calculates the weight of the feature value. In the step, for example, the weight of the feature value is more _ the input variable is difficult to calculate, and the domain recognizes whether the condition is met. If not, the process returns to step (10). In step S44, the calculation is stopped and the updated feature value is output. In summary, the invention provided by the present invention has at least the following effects as compared with the prior art: (1) The amount of non-essential data can be effectively reduced, and the problem of insufficient storage capacity can be solved. (2) The complexity of the information can be reduced to avoid slowing the transmission speed and processing resources. The above-described embodiments are merely illustrative of the principles of the invention and its effects, and are not intended to limit the invention. Modifications and alterations of the above-described embodiments can be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention should be as set forth in the scope of the patent application Π1320 13 201102941, which will be described later. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic view of a care device of the present invention; Fig. 2 is a schematic view showing a specific embodiment of the care device of the present invention; and Fig. 3 is a view showing the neural network technology of the present invention. Schematic diagram of the control part of the minimum bisector algorithm; and Fig. 4 shows a flow chart of the eigenvalue capture using the modulating partial least squares algorithm of the care device of the present invention. [Main component symbol description] 1 Care device 2 Care platform 11 Signal capture module 12, 312 Processing module 121 Operation judgment rule unit 13, 313 Communication module 3 Care ward 31 Electric wheelchair 310 Video detector 311 Physiological signal detection Detector 32 luminaire 33 alarm 4 wireless network S40~S44 5 hospital care monitoring center step 14 Π 1320