WO2021164654A1 - 时间连续性侦测判断系统及方法 - Google Patents

时间连续性侦测判断系统及方法 Download PDF

Info

Publication number
WO2021164654A1
WO2021164654A1 PCT/CN2021/076265 CN2021076265W WO2021164654A1 WO 2021164654 A1 WO2021164654 A1 WO 2021164654A1 CN 2021076265 W CN2021076265 W CN 2021076265W WO 2021164654 A1 WO2021164654 A1 WO 2021164654A1
Authority
WO
WIPO (PCT)
Prior art keywords
time
characteristic state
thermal image
judgment
state
Prior art date
Application number
PCT/CN2021/076265
Other languages
English (en)
French (fr)
Inventor
黄彦铭
Original Assignee
艾科科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 艾科科技股份有限公司 filed Critical 艾科科技股份有限公司
Publication of WO2021164654A1 publication Critical patent/WO2021164654A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H70/00—ICT specially adapted for the handling or processing of medical references
    • G16H70/60—ICT specially adapted for the handling or processing of medical references relating to pathologies

Definitions

  • the invention relates to a time continuity detection and judgment system, in particular to a system and method for judging and detecting target states by applying time-series thermal images.
  • the present invention provides a time continuity detection and judgment system, which includes: a thermal energy sensor, which continuously receives a thermal image of a detected target according to time; The thermal image is processed to obtain the first characteristic state, and the thermal image of the current time, the thermal image of the past few time points, and the characteristic state of the thermal image of the previous time point are analyzed and processed according to the second model to obtain the second characteristic state, The first characteristic state and the second characteristic state are processed according to a third model to obtain the possibility of various judgment results of the detection target at the moment.
  • the time continuity detection and judgment system further includes: a display unit to display the possibility of various judgment results of the detection target at the moment; or a wireless transmission unit to transfer the thermal energy The thermal image continuously received by the sensor according to time is transmitted to the computing unit.
  • the first model includes a convolutional neural network
  • the extracted first feature state includes a one-dimensional result and a two-dimensional result
  • the one-dimensional result includes determining the preliminary classification and whether the image is valid
  • the two-dimensional result includes a region of interest (ROI) and surrounding environment characteristics.
  • ROI region of interest
  • the second model includes a recursive neural network, and the characteristic state generated at the previous time point can be fed back to the current time point to generate a new characteristic state, and the characteristic state fed back includes continuous posture judgment actions and values. Cumulative amount.
  • the third model includes a convolutional neural network and a multilayer perceptron.
  • the present invention also provides a time continuity detection and judgment method, including the following steps: continuously receiving a thermal image of a detection target according to time; using a convolutional neural network to process the thermal image of the current time to obtain a first characteristic state; Use the recurrent neural network to analyze and process the thermal image of the current time, the thermal image of the past few time points, and the feature state of the thermal image of the previous time point to obtain the second feature state; and use the convolutional neural network and the multilayer perceptron
  • the first characteristic state and the second characteristic state are processed to obtain the possibility of various judgment results of the detection target at the moment.
  • the captured first feature state includes a one-dimensional result and a two-dimensional result, wherein the one-dimensional result includes determining whether the preliminary classification and the image are valid, and the two-dimensional result includes a field of interest (ROI). ) And the surrounding environment characteristics.
  • ROI field of interest
  • the recursive neural network can feed back the characteristic state generated at the previous time point to the current time point to generate a new characteristic state.
  • the feedback characteristic state includes continuous posture judgment actions and numerical accumulation.
  • the first characteristic state and the second characteristic state are processed by using a convolutional neural network and a multilayer perceptron, it further includes inputting background information of the detection target.
  • the detection target is a patient
  • the background information of the detection target includes diagnosis results, medication information, and chronic diseases of the patient.
  • the present invention uses the thermal energy sensor to detect time continuity, and analyzes the image characteristics of the detected target at the current time point and the behavior characteristics of the previous continuous time through the first model and the second model, and then conducts comprehensive evaluation through the third model.
  • the possibility of various situations of the detection target is generated.
  • it can also further analyze the overall behavior of the detection target, and provide medical staff as a reference for follow-up care or treatment.
  • FIG. 1 shows the architecture diagram of a time continuity detection and judgment system provided by the present invention.
  • Figure 2 shows the operation flow performed by the arithmetic unit.
  • FIG. 3 shows a flowchart of a time continuity detection and judgment method provided by the present invention.
  • the first model 21 is the first model 21
  • FIG. 1 shows the architecture of a time continuity detection and judgment system 1 provided by the present invention.
  • the time continuity detection and judgment system 1 can be applied to detect and judge human actions, such as getting up, getting out of bed, falling, turning over, activity volume, and can be extended to overall behavior judgment, such as long time in bed, sleep quality, etc. , But not limited to this.
  • the components of the time continuity detection and judgment system 1 mainly include a thermal energy sensor 11 and an arithmetic unit 12.
  • the thermal sensor 11 can continuously receive thermal images of a detection target (for example, a patient) according to time.
  • a detection target for example, a patient
  • the first model 21 includes a convolutional neural network
  • the extracted first feature state 210 includes a one-dimensional result and a two-dimensional result, wherein the one-dimensional result includes determining the preliminary classification and whether the image is Effective, the two-dimensional result includes a region of interest (ROI) and surrounding environment characteristics.
  • ROI region of interest
  • the characteristic state of feedback includes continuous posture judgment action and numerical accumulation. That is, the second model 22 can combine the thermal image content at the previous several time points and the features captured at the previous time point to include sequential action features.
  • the value and the result of the judgment at each time point are formed by the state machine calculation to form a continuous action judgment and so on.
  • the third model 23 includes a convolutional neural network and a multi-layer perceptron, and by combining single-time and continuous-time coupled data, various judgments that ultimately fit the current situation can be formed. In addition, since the continuous action is highly correlated with the current situation, the third model 23 can effectively use the coupling to converge the result. In this way, by combining the first characteristic state 210 and the second characteristic state 220 through the third model 23, the possibility of different situations can be analyzed, and the possibility 230 of various current judgment results can be inferred. Furthermore, the third model 23 receives and couples the output of the first model 21 and the second model 22 and time information.
  • the output of the third model 23 it is the possibility of several action judgments, such as turning over, getting out of bed, and falling. And several arithmetic values, such as activity score, sleep quality score, bedtime, etc.
  • the analysis in the model uses each input classification and then calculates according to the weight.
  • the time continuity detection and judgment system 1 may optionally include a display unit 13 or a wireless transmission unit 14 and a power management unit 15.
  • the display unit 13 is used to display the possibility 230 of various judgment results at the moment.
  • the wireless transmission unit 14 is an optional component for transmitting the thermal image received by the thermal energy sensor 11 to the computing unit 12.
  • the power management unit 15 can provide power required for the operation of the thermal energy sensor 11, the computing unit 12, the display unit 13, or the wireless transmission unit 14 according to actual needs.
  • the application of the time continuity detection and judgment system 1 does not limit the construction of the above-mentioned components in a single device, but may be constructed on several different devices.
  • the computing unit 12 may be set in the cloud or on a server, and the thermal energy sensor 11 may be set in the ward to receive thermal images of the detected target (for example, a patient), and upload it to the computing unit 12 through the wireless transmission unit 14.
  • the display unit 13 may be a monitoring screen of a nursing station or a screen of a handheld device of a medical staff to display the result of the calculation by the calculation unit 12.
  • the present invention also provides a time continuity detection and judgment method, including the following steps:
  • the first characteristic state captured in step S12 includes a one-dimensional result and a two-dimensional result, wherein the one-dimensional result includes determining whether the preliminary classification and the image are valid, and the two-dimensional result includes attention ROI and surrounding environment characteristics.
  • the characteristic state taught includes continuous posture judgment actions and numerical accumulation.
  • the detection target when using a convolutional neural network and a multilayer perceptron to process the first characteristic state and the second characteristic state in step S14, it further includes inputting background information of the detection target.
  • background information may include diagnosis results, medication information, and related data that affect the patient's dynamics.
  • the present invention uses the thermal energy sensor to detect time continuity, and analyzes the image characteristics of the detected target at the current time point and the behavior characteristics of the previous continuous time through the first model and the second model, and then conducts comprehensive evaluation through the third model.
  • the possibility of various situations of the detection target is generated.
  • it can also further analyze the overall behavior of the detection target, and provide medical staff as a reference for follow-up care or treatment.
  • a bedridden patient judges that he is lying down and still during the day, and has no movement for a continuous period of time, judges no turning over, has low activity, and has been in bed for a long time, and bed sores are prone to occur; another example: the patient judges in the afternoon The current state is a falling posture.
  • the continuous time characteristic is that you are preparing to get out of bed. If you are judged to be falling while getting out of bed, you must notify the relevant personnel; another example: the user judges that the current is turned over at night, and the continuous time feature includes getting out of bed, judging the user’s sleep quality. good.
  • the examination and judgment system provided by the present invention can assist medical staff to understand the patient's state more clearly.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Primary Health Care (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Epidemiology (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Image Analysis (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

一种时间连续性侦测判断系统，包括：热能传感器，依时间连续接收一侦测目标的热影像；及运算单元，根据第一模型对当下时间的热影像进行处理以取得第一特征状态，根据第二模型对当下时间的热影像、过去数时间点前的热影像以及前一个时间点热影像的特征状态进行分析处理以取得第二特征状态，根据第三模型对第一特征状态与第二特征状态进行处理，以得到侦测目标当下各种判断结果的可能性。

Description

时间连续性侦测判断系统及方法 技术领域
本发明涉及一种时间连续性侦测判断系统，尤指一种应用时序性热影像判断侦测目标状态的系统及方法。
背景技术
现有技术中，为了减轻医护人员的负担，提供医生作更佳的判断辅助来源，以及让病患受到更妥善照顾，各式各样的穿戴式装置，以至于侵入式装置，大量应用于医疗系统中，以随时侦测病患的状况。不过，大多数的装置皆是针对生理数据进行监测，例如:心跳、血氧、体温、血压等等。部份装置虽然也具有动作感知的功能，不过其普遍只关注于患者当下的动作或状态，而缺乏对所感测的数据进一步的分析。
因此，如果能开发一种侦测患者动作状态的工具，并且可从陆续收集到的数据中，分析判断出患者当下的动作与整体行为的可能肇因，则可提供医护人员作为后续照护或处置的参考依据，显然能大幅提升医护质量。
发明内容
为解决现有技术的问题，本发明提供一种时间连续性侦测判断系统，包括：热能传感器，依时间连续接收一侦测目标的热影像；及运算单元，根据第一模型对当下时间的热影像进行处理以取得第一特征状态，根据第二模型对当下时间的 热影像、过去数时间点前的热影像以及前一个时间点热影像的特征状态进行分析处理以取得第二特征状态，根据第三模型对所述第一特征状态与所述第二特征状态进行处理，以得到所述侦测目标当下各种判断结果的可能性。
在一实施例中，所述时间连续性侦测判断系统，更包括:显示单元，用以显示所述侦测目标当下各种判断结果的可能性；或无线传输单元，用以将所述热能传感器依时间连续接收的热影像传送至所述运算单元。
在一实施例中，所述第一模型包括卷积神经网络，所撷取所述第一特征状态包含一维结果与二维结果，其中所述一维结果包含判断初步分级以及影像是否有效，所述二维结果包含关注区域(ROI)与周围环境特征。
在一实施例中，所述第二模型包括递归神经网络，可由前一时间点产生的特征状态回授至当下时间点以产生新的特征状态，回授的特征状态包含连续姿态判断动作、数值累计量。
在一实施例中，所述第三模型包括卷积神经网络与多层感知器。
本发明也提供一种时间连续性侦测判断方法，包括下列步骤：依时间连续接收一侦测目标的热影像；使用卷积神经网络对当下时间的热影像进行处理以取得第一特征状态；使用递归神经网络对当下时间的热影像、过去数时间点前的热影像以及前一个时间点热影像的特征状态进行分析处理以取得第二特征状态；及使用卷积神经网络与多层感知器对所述第一特征状态与所述第二特征状态进行处理，以得到所述侦测目标当下各种判断结果的可能性。
在一实施例中，所撷取所述第一特征状态包含一维结果与二维结果，其中所述一维结果包含判断初步分级以及影像是否有效，所述二维结果包含关注场域(ROI)与周围环境特征。
在一实施例中，所述递归神经网络可将前一时间点产生的特征状态回授至当下时间点以产生新的特征状态，回授的特征状态包含连续姿态判断动作、数值累计量。
在一实施例中，所述使用卷积神经网络与多层感知器对所述第一特征状态与所述第二特征状态进行处理时，更包括输入所述侦测目标的背景信息。其中当所述侦测目标为一患者时，所述侦测目标的背景信息包含所述患者的诊断结果、用药信息、痼疾。
本发明利用热能传感器进行时间连续性的侦测，并通过第一模型与第二模型分析侦测目标当下时间点的影像特征以及先前连续时间的行为特征，再通过第三模型并进行综合评估，而产生所述侦测目标各种情况的可能性。除了可提供辨别侦测目标当下的动作状态外，尚可进一步解析侦测目标的整体行为，而可提供医护人员作为后续照护或处置的参考依据。
附图说明
图1显示本发明所提供一种时间连续性侦测判断系统的架构图。
图2显示运算单元所执行的运算流程。
图3显示本发明所提供一种时间连续性侦测判断方法的流程图。
时间连续性侦测判断系统    1
热能传感器      11
运算单元        12
显示单元        13
无线传输单元    14
电源管理单元    15
第一模型        21
第一特征状态    210
第二模型        22
第二特征状态    220
第三模型        23
当下各种判断结果的可能性    230
步骤           S11～S14
具体实施方式
以下藉由特定的具体实施例说明本发明的实施方式，熟悉此技术的人士可由本说明书所揭示的内容轻易地了解本发明的其他优点与功效。本发明亦可藉由其他不同的具体实施例加以施行或应用。
须知，本说明书所附图式所绘示的结构、比例、大小等，均仅用以配合说明书所揭示的内容，以供熟悉此技艺的人士的了解与阅读，并非用以限定本发明可实施的限定条件，故不具技术上的实质意义，任何结构的修饰、比例关的改变或大小的调整，在不影响本发明所能产生的功效及所能达成的目的下，均应落在本发明所揭示的技术内容得能涵盖的范围内。同时，本说明书中所引用的如「上」、「内」、「外」、「底」及「一」等的用语，亦仅为便于叙述的明了，而非用以限定本发明可实施的范围，其相对关的改变或调整，在无实质变更技术内容下，当亦视为本发明可实施的范畴，合先叙明。
请参阅图1，显示本发明所提供一种时间连续性侦测判断系统1的架构。所述时间连续性侦测判断系统1可应用于侦测判断人体动作，例如:起床、下床、 跌倒、翻身、活动量，并可扩及整体行为判断，例如:长时间卧床、睡眠质量等，但不限于此。在一实施例中，所述时间连续性侦测判断系统1的组件主要包括了一热能传感器11以及一运算单元12。
热能传感器11可依时间连续接收一侦测目标(例如:患者)的热影像。至于运算单元12所执行的运算流程，请参阅图2，其可根据第一模型21对当下时间(t＝n)的热影像进行处理以取得第一特征状态210。同时，根据第二模型22对当下时间(t＝n)的热影像、过去数时间点前(t＝n-k)的热影像以及前一个时间点(t＝n-1)热影像的特征状态进行分析处理以取得第二特征状态220。接着，根据第三模型23对第一特征状态210与第二特征状态220进行合并处理，以分析出不同情况的可能性，最终推论得到侦测目标当下时间(t＝n)各种判断结果的可能性230。
在一实施例中，所述第一模型21包括卷积神经网络，所撷取所述第一特征状态210包含一维结果与二维结果，其中所述一维结果包含判断初步分级以及影像是否有效，所述二维结果包含关注区域(ROI)与周围环境特征。
在一实施例中，所述第二模型22包括递归神经网络，可由前一时间点(t＝n-1)产生的特征状态回授至当下时间点(t＝n)以产生新的特征状态，回授的特征状态包含连续姿态判断动作、数值累计量。亦即，第二模型22可结合先前数时间点的热影像内容以及前一时间点所撷取特征而包含时序性动作特征。更进一步说，第二模型22可输入过去数时间点前(t＝n-k)的热影像数据堆栈而成的三维数据，以及前一个时间点(t＝n-1)的回授特征，至于第二模型22的输出则为过去数时间点前(t＝n-k)以及前一个时间点(t＝n-1)的撷取特征，包括中心区块及其周围大小不一的关注区域、全局极值、各时间点形成判断的结果通过状态机运算形成连续动作判断等。
在一实施例中，所述第三模型23包括卷积神经网络与多层感知器，藉由结合单一时间与连续时间耦合的数据，而能形成最终符合当下情境的各种判断。并且，由于连续动作与当下情境高度相关，因此第三模型23能有效利用耦合而收敛结果。藉此，通过第三模型23将第一特征状态210与第二特征状态220合并考虑，可分析出不同情况的可能性，并推论得到当下各种判断结果的可能性230。进一步说，第三模型23接收并耦合第一模型21与第二模型22的输出以及时间信息，至于第三模型23的输出则为数个动作判断的可能性，例如：翻身、下床、跌倒，以及数个运算数值，例如：活动量分数、睡眠质量分数、卧床时间等，模型内分析采用各输入分类后依照权重计算而得。
仍请参阅图1，除了前述热能传感器11与运算单元12外，所述时间连续性侦测判断系统1并可选择性地包括一显示单元13或一无线传输单元14以及一电源管理单元15。显示单元13，用以显示当下各种判断结果的可能性230。无线传输单元14，则为一选择性组件，用以将热能传感器11所接收的热影像传送至运算单元12。电源管理单元15可根据实际需求提供热能传感器11、运算单元12、显示单元13或无线传输单元14操作所需的电力。
值得注意的是，本发明所提供的时间连续性侦测判断系统1在应用上，并不限定将上述组成单元建构于单一装置中，而可能建构于数个不同的装置上。例如，运算单元12可能设置在云端或服务器上，热能传感器11可能设置于病房中以接收侦测目标(例如:患者)的热影像，并且通过无线传输单元14上传至运算单元12，至于显示单元13则可能为护理站的监控屏幕或是医护人员手持装置的屏幕，用以显示运算单元12运算后的结果。
请参阅图3，本发明并提供一种时间连续性侦测判断方法，包括下列步骤：
S11:依时间连续接收一侦测目标的热影像；
S12:使用卷积神经网络对当下时间(t＝n)的热影像进行处理以取得第一特征状态；
S13:使用递归神经网络对当下时间(t＝n)的热影像、过去数时间点前(t＝n-k)的热影像以及前一个时间点(t＝n-1)热影像的特征状态进行分析处理以取得第二特征状态；及
S14:使用卷积神经网络与多层感知器对第一特征状态与第二特征状态进行合并处理，以分析出不同情况的可能性，最终推论得到侦测目标当下时间(t＝n)各种判断结果的可能性。
在一实施例中，于步骤S12所撷取的所述第一特征状态包含一维结果与二维结果，其中所述一维结果包含判断初步分级以及影像是否有效，所述二维结果包含关注场域(ROI)与周围环境特征。
在一实施例中，于步骤S13中的递归神经网络可将前一时间点(t＝n-1)产生的特征状态回授至当下时间点(t＝n)以产生新的特征状态，回授的特征状态包含连续姿态判断动作、数值累计量。
在一实施例中，于步骤S14中使用卷积神经网络与多层感知器对所述第一特征状态与所述第二特征状态进行处理时，更包括输入所述侦测目标的背景信息。例如:当侦测目标为一患者时，其背景信息可包含诊断结果、用药信息、痼疾等会影响病患动态的相关数据。
本发明利用热能传感器进行时间连续性的侦测，并通过第一模型与第二模型分析侦测目标当下时间点的影像特征以及先前连续时间的行为特征，再通过第三模型并进行综合评估，而产生所述侦测目标各种情况的可能性。除了可提供辨别 侦测目标当下的动作状态外，尚可进一步解析侦测目标的整体行为，而可提供医护人员作为后续照护或处置的参考依据。在一些实施例中，例如:卧床患者在白天判断当下躺卧静止，且连续时间皆无动作，判断无翻身、活动量低、且卧床时间长，易有褥疮情事发生；又如:患者午后判断当下为跌倒姿态，连续时间特征为正准备下床，判断为下床时跌倒，须通知相关人员；再如:使用者夜间判断当下为翻身状态，连续时间特征包含下床，判断用户睡眠质量不佳。显然，通过本发明所提供的检查判断系统，可以协助医护人员更明确的了解患者的状态。
藉由以上较佳具体实施例的描述，本领域具有通常知识者当可更加清楚本发明的特征与精神，惟上述实施例仅为说明本发明的原理及其功效，而非用以限制本发明。因此，任何对上述实施例进行的修改及变化仍不脱离本发明的精神，且本发明的权利范围应如权利要求书所列。

Claims (10)

  1. 一种时间连续性侦测判断系统，其特征在于，所述系统包括：
    热能传感器，依时间连续接收一侦测目标的热影像；及
    运算单元，根据第一模型对当下时间的热影像进行处理以取得第一特征状态，根据第二模型对当下时间的热影像、过去数时间点前的热影像以及前一个时间点热影像的特征状态进行分析处理以取得第二特征状态，根据第三模型对所述第一特征状态与所述第二特征状态进行处理，以得到所述侦测目标当下各种判断结果的可能性。
  2. 如权利要求1所述的时间连续性侦测判断系统，其特征在于，所述系统更包括:
    显示单元，用以显示所述侦测目标当下各种判断结果的可能性；或
    无线传输单元，用以将所述热能传感器依时间连续接收的热影像传送至所述运算单元。
  3. 如权利要求1所述的时间连续性侦测判断系统，其特征在于，所述第一模型包括卷积神经网络，所撷取所述第一特征状态包含一维结果与二维结果，其中所述一维结果包含判断初步分级以及影像是否有效，所述二维结果包含关注区域(ROI)与周围环境特征。
  4. 如权利要求1所述的时间连续性侦测判断系统，其特征在于，所述第二模型包括递归神经网络，可由前一时间点产生的特征状态回授至当下时间点以产生新的特征状态，回授的特征状态包含连续姿态判断动作、数值累计量。
  5. 如权利要求1所述的时间连续性侦测判断系统，其特征在于，所述第三模型包括卷积神经网络与多层感知器。
  6. 一种时间连续性侦测判断方法，其特征在于，所述方法包括下列步骤：
    依时间连续接收一侦测目标的热影像；
    使用卷积神经网络对当下时间的热影像进行处理以取得第一特征状态；
    使用递归神经网络对当下时间的热影像、过去数时间点前的热影像以及前一个时间点热影像的特征状态进行分析处理以取得第二特征状态；及
    使用卷积神经网络与多层感知器对所述第一特征状态与所述第二特征状态进行处理，以得到所述侦测目标当下各种判断结果的可能性。
  7. 如权利要求6所述的时间连续性侦测判断方法，其特征在于，撷取所述第一特征状态包含一维结果与二维结果，其中所述一维结果包含判断初步分级以及影像是否有效，所述二维结果包含关注场域(ROI)与周围环境特征。
  8. 如权利要求6所述的时间连续性侦测判断方法，其特征在于，所述递归神经网络可将前一时间点产生的特征状态回授至当下时间点以产生新的特征状态，回授的特征状态包含连续姿态判断动作、数值累计量。
  9. 如权利要求6所述的时间连续性侦测判断方法，其特征在于，所述使用卷积神经网络与多层感知器对所述第一特征状态与所述第二特征状态进行处理时，更包括输入所述侦测目标的背景信息。
  10. 如权利要求9所述的时间连续性侦测判断方法，其特征在于，当所述侦测目标为一患者时，所述侦测目标的背景信息包含所述患者的诊断结果、用药信息、痼疾。
PCT/CN2021/076265 2020-02-20 2021-02-09 时间连续性侦测判断系统及方法 WO2021164654A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010105922.4A CN113284607A (zh) 2020-02-20 2020-02-20 时间连续性侦测判断系统及方法
CN202010105922.4 2020-02-20

Publications (1)

Publication Number Publication Date
WO2021164654A1 true WO2021164654A1 (zh) 2021-08-26

Family

ID=77275081

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/076265 WO2021164654A1 (zh) 2020-02-20 2021-02-09 时间连续性侦测判断系统及方法

Country Status (2)

Country Link
CN (1) CN113284607A (zh)
WO (1) WO2021164654A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102387345A (zh) * 2011-09-09 2012-03-21 浙江工业大学 基于全方位视觉的独居老人安全监护系统
CN103955699A (zh) * 2014-03-31 2014-07-30 北京邮电大学 一种基于监控视频的实时摔倒事件检测方法
US20150199892A1 (en) * 2003-12-12 2015-07-16 Careview Communications, Inc Noise Correcting Patient Fall Risk State System and Method for Predicting Patient Falls
CN107945458A (zh) * 2017-12-11 2018-04-20 北京奇虎科技有限公司 摄像头场景下的跌倒检测方法、装置及计算设备
CN108629300A (zh) * 2018-04-24 2018-10-09 北京科技大学 一种跌倒检测方法
CN108960056A (zh) * 2018-05-30 2018-12-07 西南交通大学 一种基于姿态分析和支持向量数据描述的跌倒检测方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150199892A1 (en) * 2003-12-12 2015-07-16 Careview Communications, Inc Noise Correcting Patient Fall Risk State System and Method for Predicting Patient Falls
CN102387345A (zh) * 2011-09-09 2012-03-21 浙江工业大学 基于全方位视觉的独居老人安全监护系统
CN103955699A (zh) * 2014-03-31 2014-07-30 北京邮电大学 一种基于监控视频的实时摔倒事件检测方法
CN107945458A (zh) * 2017-12-11 2018-04-20 北京奇虎科技有限公司 摄像头场景下的跌倒检测方法、装置及计算设备
CN108629300A (zh) * 2018-04-24 2018-10-09 北京科技大学 一种跌倒检测方法
CN108960056A (zh) * 2018-05-30 2018-12-07 西南交通大学 一种基于姿态分析和支持向量数据描述的跌倒检测方法

Also Published As

Publication number Publication date
CN113284607A (zh) 2021-08-20

Similar Documents

Publication Publication Date Title
Sangeethalakshmi et al. Patient health monitoring system using IoT
US7420472B2 (en) Patient monitoring apparatus
US7502498B2 (en) Patient monitoring apparatus
US20200237225A1 (en) Wearable patient monitoring systems and associated devices, systems, and methods
CN108899084A (zh) 一种智慧养老健康监护系统
US10665332B2 (en) Systems and methods for facilitating physiological data collection prior to appointment
Kim et al. Ubiquitous healthcare system for analysis of chronic patients’ biological and lifelog data
WO2022077643A1 (zh) 一种用于监视危重病人的护理机器人
TWM562100U (zh) 智能檢測設備
WO2021164654A1 (zh) 时间连续性侦测判断系统及方法
Nerella et al. AI-enhanced intensive care unit: revolutionizing patient care with pervasive sensing
CN108652621A (zh) 一种吞咽实时监测方法及其系统
TWI772762B (zh) 照護系統與自動照護方法
TWI741502B (zh) 時間連續性偵測判斷系統及方法
JP2006116325A (ja) 状態解析装置
Farah et al. Respiratory rate measurement in children using a thermal imaging camera
TW201102941A (en) Caretaking device
TWM582674U (zh) Care system
WO2021182455A1 (ja) 情報処理方法、コンピュータプログラム、情報処理装置及び情報処理システム
Al-Kalidi et al. Respiratory rate measurement in children using a thermal camera
Sheikdavood et al. Smart Health Monitoring System for Coma Patients using IoT
Palash et al. Low-cost smart COVID-19 patient monitoring and support system
US20230074628A1 (en) System and method for automating bedside infection audits using machine learning
Selvaraj et al. AN INNOVATION TECHNIQUE OF ARTIFICIAL INTELLIGENCE ENABLED DOCTOR ASSISTANT IN FIELD OF HEALTH CARE
Sunil et al. IOT based Smart Health Analyzer

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21757841

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21757841

Country of ref document: EP

Kind code of ref document: A1