WO2016091102A1 - Fuzzy neural network-based body area network health information monitoring and interaction system - Google Patents
Fuzzy neural network-based body area network health information monitoring and interaction system Download PDFInfo
- Publication number
- WO2016091102A1 WO2016091102A1 PCT/CN2015/096198 CN2015096198W WO2016091102A1 WO 2016091102 A1 WO2016091102 A1 WO 2016091102A1 CN 2015096198 W CN2015096198 W CN 2015096198W WO 2016091102 A1 WO2016091102 A1 WO 2016091102A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- module
- layer
- fuzzy
- data
- output
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Definitions
- the present invention relates to a body network health information monitoring and interaction system based on a fuzzy neural network.
- the existing medical health platform has the following problems:
- the present invention proposes a body network health information monitoring based on fuzzy neural network. Measuring and interacting system, the system uses fuzzy neural network algorithm to upload the user's health information in time, share and sort in the specified range, and make everyone pay attention to their health, and check regularly.
- a body network health information monitoring and interaction system based on a fuzzy neural network comprising a collecting end, a fuzzy neural network analyzing and detecting system, an interaction module, an identifying unit and a display unit;
- the collection end includes an acquisition module and a control module, and the collection module is configured to collect related biological data collected by the home medical device, and send the data to the control module, where the control module is further configured to control the collection module to collect data;
- the interaction module is configured to establish an interaction group, select an interaction group size, a number of collection ends that perform data interaction with each other, and establish data sharing;
- the fuzzy neural network analysis and detection system comprises a model building module, a fuzzy neural network data analysis module and a sorting module, wherein:
- the model construction module receives the biological data transmitted in the interaction group in the interaction module, and uses it as an input variable to construct a membership function of the biological data through the input layer and the fuzzy layer module;
- the fuzzy neural network data analysis module is configured to process the biological data by constructing a fuzzy inference layer, a layer back layer, a deblurring layer, and an output layer module, and calculate an output biological index value;
- the sorting module is configured to perform data ranking in the interaction group according to the biometric index value
- the identification unit receives the data transmitted by the sorting module, reads the interaction group of the interaction module, identifies the collection module and the display unit in the same interaction group, and displays the biological data and the data ranking through the display unit.
- the acquisition module monitors data uploaded by the home medical device in real time, including systolic blood pressure, diastolic blood pressure, and heart rate.
- the model building module includes an input layer module and a fuzzification layer module, wherein the input layer module is a first layer module, and the input value is data uploaded after measurement by the home medical device, and the uploaded data includes shrinkage.
- the fuzzification layer module is a second layer module, and the input data is divided into a fuzzy subset, and the fuzzy subset of the systolic pressure SSY, the diastolic-pressure SZY, and the heart rate XL is ⁇ normal (M), high (H) ), very high (VH) ⁇ , membership degrees are SSY (M, H, VH), ⁇ SZY (M, H, VH) and XL (M, H, VH).
- the fuzzification layer 2 module for "normal (M)", uses a symmetric function of the Sigmoid function, and the membership degree is:
- Wl, W2 are systems, the initial value is arbitrary, and the network training and self-learning are continuously adjusted to minimize the error between the actual output value of the network and the tutor signal.
- the node output range of this layer is between 0 and 1.
- the final output of the second layer module is:
- the fuzzy neural network data analysis module includes a fuzzy inference layer module, which is set as a third layer module, and is used to establish a fuzzy inference rule base, wherein the language rule is a basis of fuzzy logic reasoning, and each fuzzy inference rule is Conclusion
- the fuzzy neural network data analysis module includes a fourth layer module, that is, a layer one module, and the total number of nodes is the same as that of the third layer module, realizing normalization processing on the fourth layer output, preparing for deblurring , its output is:
- the fuzzy neural network data analysis module further includes a deblurring layer module, configured to obtain, by fuzzy inference, membership degrees of different rules corresponding to different outputs (H, SubH, UnH), by deblurring
- the method obtains the membership of three health conditions, and the output of this layer is:
- the application of the weight W nl optimizes the fuzzy inference rule and makes the fuzzy reasoning adaptive, thereby improving the intelligence degree of the fuzzy neural network model.
- the relative importance of the subset (H, SubH, UnH), taking W That is, the output of the fuzzy subset (H) contributes to the health index value of 1, the sub-health (SubH) contributes to the health index value of 0.5, and the unhealthy (UnH) contributes to the health index value of 0;
- the health index value P is:
- the fuzzy neural network data analysis module further includes a BP neural network learning training module, configured to train the network by using an improved BP algorithm with an additional momentum term, and adjust the weight by using a gradient descent algorithm of error back propagation.
- the weight in the objective function gradient along the reverse direction is adjusted to confirm the fuzzy neural network parameters Wl, W2 and W nl, first adjust W nl, then adjust Wl, W2.
- the total mean square error of the actual output of the network and the expected output is minimized.
- P(n) is the actual output of the nth iteration and P is the desired output.
- the purpose of learning training is to minimize the error function.
- ⁇ is the learning rate, 0 ⁇ 1, n AW n)
- the value is 0.9.
- the sorting module reads an output value of the output layer module, reads an acquisition module corresponding to the output value, and confirms that the collection module collects other outputs in the interaction group through the interaction group selected by the control module in the interaction module.
- the output value of the layer module, and the data is ranked within the interaction group.
- each of the medical device outputs in the body area network
- the health data is sent to the fuzzy neural network system model for analysis, and the health index value is obtained, and the health ranking is performed, which can stimulate the enthusiasm of the user, and let them pay attention to their own health problems in time, and can use the home medical equipment to measure on time;
- the system has good scalability, and can join more other types of health data for interaction.
- FIG. 1 is a schematic structural view of the present invention.
- FIG. 2 is a fuzzy neural network model diagram of the present invention.
- 3 is a graph of membership function of the present invention.
- a body network health information monitoring and interaction system based on a fuzzy neural network includes a collection end, a fuzzy neural network analysis and detection system, an interaction module, an identification unit, and a display unit.
- the collection end includes an acquisition module and a control module, and the collection module is configured to collect related biological data collected by the home medical device, and send the data to the control module, where the control module is further configured to control the collection module to collect data;
- the interaction module is configured to establish an interaction group, select an interaction group size, a number of collection ends that perform data interaction with each other, and establish data sharing;
- the fuzzy neural network analysis and detection system comprises a model construction module, a fuzzy neural network data analysis module and a sorting module, wherein:
- the model building module receives the biological data transmitted in the interaction group in the interaction module, and uses it as an input variable to construct a membership function of the biological data through the input layer and the fuzzification layer module;
- the fuzzy neural network data analysis module is configured to process the biological data by constructing a fuzzy inference layer, a layer back layer, a deblurring layer, and an output layer module, and calculate an output biological index value;
- the sorting module is configured to perform data ranking in the interaction group according to the biometric index value
- the identification unit receives the data transmitted by the sorting module, reads the interaction group of the interaction module, identifies the collection module and the display unit in the same interaction group, and ranks the biological data and the data through the display unit. Do not.
- the collection module monitors data uploaded by the home medical device in real time, including systolic blood pressure, diastolic blood pressure, and heart rate.
- the fuzzy neural network model includes an input layer, a fuzzification layer, a fuzzy inference layer, a normalization layer, a deblurring layer, and an output layer, as shown in FIG. 2.
- the third layer is the fuzzy inference layer, and each node represents a fuzzy rule. Therefore, all Layer 3 nodes constitute a fuzzy rule base.
- the connection between the second layer and the third layer acts as a connection inference engine, avoiding the fuzzy implication reasoning process.
- the second layer of the connection defines the predecessor of the fuzzy rule
- the third layer of the connection defines the posterior of the fuzzy rule.
- the normal direction of propagation is from the first layer to the sixth layer, that is, from bottom to top.
- the signal propagates from the sixth layer to the first layer, that is, from top to bottom.
- the input and output relationship of each layer is as follows (0 (x) indicates the output of the Xth layer).
- the first layer is an input layer.
- the input value is the data uploaded after measurement by the home medical device.
- the uploaded data includes systolic blood pressure SSY, diastolic blood pressure SZY and heart rate XL.
- This layer only serves as a pass, passing the input data to the next layer.
- the output is:
- the second layer is a fuzzification layer.
- This layer divides the input data into fuzzy subsets.
- the fuzzy subsets of systolic pressure SSY, diastolic pressure SZY and heart rate XL are ⁇ normal (M), high (H), very high (VH) ⁇ , and the membership is W SSi. (M, H, VH), SZY (M, H, VH) and ⁇ ⁇ ( ⁇ , ⁇ , ⁇ ).
- M, H, VH SZY
- ⁇ ⁇ ⁇ , ⁇ , ⁇
- the tutor's signal error is minimal, and the adjusted weight can be used for actual health assessment.
- the node output range is between 0.
- the role of the parameter W2 is to shift the membership function to the right in the horizontal axis; the role of W1 is to adjust the shape of the membership function, and the larger W1 makes the function approximate the step. W1 makes the function flatter.
- W2 and W1 represent the center and width of the membership function, respectively. The closer the input sample is to the center of the node, the larger the output.
- the final output of the second layer is:
- the third layer is a fuzzy inference layer.
- the applicability of each inference rule a (l) is:
- fuzzy rules are expressed in terms of linguistic variables:
- C is the probability of being in a certain state of health.
- the output of the third layer is:
- the fourth layer is a normalized layer.
- the total number of nodes is the same as that of the third layer, and the normalized processing of the fourth layer output is implemented to prepare for the deblurring.
- the output is:
- the fifth layer is a deblurring layer.
- the fuzzy inference results in the membership of different rules corresponding to different outputs (H, SubH, Un H), and the degree of membership of the three health conditions needs to be obtained by deblurring.
- the output of this layer is:
- the weight W nl is introduced, and the relative importance of each fuzzy inference rule is discriminated by its change in the network.
- the application of weight W nl optimizes the fuzzy inference rules and makes the fuzzy reasoning adaptive, thus improving the intelligence of the fuzzy neural network model.
- the sixth layer is an output layer.
- the contribution in the fuzzy subset (H) to the health index value is 1, and the contribution of sub-health (SubH) to the health index value is 0.5, and the unhealthy (UnH) contribution to the health index value is zero.
- Its final output health index value P is:
- a learning training process of the BP neural network is also included.
- the parameters that need to be determined in the fuzzy neural network are Wl, W2, and W nl .
- the network is trained using an improved BP algorithm with additional momentum terms.
- the gradient descent algorithm of error back propagation is used to adjust the weight, so that the weight is adjusted in the opposite direction of the gradient of the objective function. Therefore, W nl is adjusted first, then Wl and W2 are adjusted. Ultimately, the total mean square error of the actual output of the network and the expected output is minimized.
- the specific network learning algorithm operation process is as follows
- the purpose of learning training is to minimize the error function.
- the input variables may also be blood oxygenation values, blood glucose values, and the like.
- the sorting module reads an output value of the output layer module, reads an acquisition module corresponding to the output value, and confirms that the collection module collects other outputs in the interaction group through the interaction group selected by the control module in the interaction module.
- the output value of the layer module, and the data is ranked within the interaction group.
Abstract
A fuzzy neural network-based body area network health information monitoring and interaction system comprises collection ends, a fuzzy neural network analysis and detection system, an interaction module, a recognition unit and a display unit; the collection end comprises a collection module and a control module; the collection module is used for collecting related biological data collected by household medical equipment and sending the data to the control module; the control module is further used for controlling the collection module to collect data; the interaction module is used for building an interactive group, selecting the size of the interactive group and the number of the collection ends for mutual data interaction, and creating data sharing; the fuzzy neural network analysis and detection system comprises a model building module, a fuzzy neural network data analysis module and a ranking module. Timely uploading of health monitoring data and interaction within a selectable range are achieved, and information interaction is achieved on the basis of protecting the privacy of users and respecting the wishes of users.
Description
说明书 Instruction manual
发明名称: 一种基于模糊神经网络的体域网健康信息监测、 交互系 统 Title: A Body Network Health Information Monitoring and Interaction System Based on Fuzzy Neural Network
技术领域 Technical field
[0001] 本发明涉及一种基于模糊神经网络的体域网健康信息监测、 交互系统。 [0001] The present invention relates to a body network health information monitoring and interaction system based on a fuzzy neural network.
背景技术 Background technique
[0002] 如今人们生活水平逐渐提高, 家用医疗设备也开始普遍使用, 但测试结果只有 本人或家属知道, 如果遇见突发情况, 没有家属等陪同人员, 医疗工作者无法 及时了解病人的身体状态、 有无病史等信息, 不能很好进行诊断、 及时治疗。 因此, 家用医疗设备的监测数据通过医疗健康平台进行共享、 存储是十分必要 的。 [0002] Nowadays people's living standards are gradually improving, and home medical equipment is also widely used, but the test results are only known to the person or family. If there is an emergency, no family members or other accompanying personnel, the medical workers cannot know the patient's physical condition in time. There is no information such as medical history, and it is not easy to diagnose and treat in time. Therefore, monitoring data of home medical equipment is necessary for sharing and storage through the health care platform.
发明概述 Summary of invention
技术问题 technical problem
[0003] 现有的医疗健康平台存在以下问题: [0003] The existing medical health platform has the following problems:
[0004] ( 1 ) 信息多为一对一交互, 即只有医生和上传者自己能够看到, 许多时候, 在用户同意的情况下, 信息也无法进行共享; 这样存在着许多隐性、 家族遗传 的问题无法发现, 不利于医生了解病人情况; [0004] (1) Information is mostly one-to-one interaction, that is, only the doctor and the uploader can see it themselves. In many cases, the information cannot be shared with the user's consent; there are many recessive, family inheritances. The problem cannot be discovered, which is not conducive to the doctor's understanding of the patient's condition;
[0005] ( 2 ) 检查的数据太过专业, 普通用户无法真正了解自身的检查情况是否健康 或存在问题, 不能生成统一的指标供用户参考; [0005] ( 2) The data to be checked is too professional. Ordinary users cannot really understand whether their inspections are healthy or have problems, and cannot generate uniform indicators for users to refer to;
[0006] ( 3 ) 同一个单位、 公司的员工的身体状况, 领导、 同事无法了解, 在出差、 体力劳动等工作的安排上可能会出现误判, 威胁身体健康; 某些关系好的用户 无法知道自己的健康程度在朋友之间排名如何, 无法激发用户的积极性, 忽略 关注自身健康问题, 不能按时使用家用医疗设备进行测量。 [0006] (3) The physical condition of the employees of the same unit or company, leaders and colleagues cannot understand that there may be misjudgments in the arrangement of work such as business trips and manual labor, which threaten the health of the body; Knowing how well your health ranks among your friends can't motivate users, ignore concerns about their health, and don't use home medical equipment to measure on time.
问题的解决方案 Problem solution
技术解决方案 Technical solution
[0007] 本发明为了解决上述问题, 提出了一种基于模糊神经网络的体域网健康信息监
测、 交互系统, 该系统利用模糊神经网络算法, 对用户的健康信息及时上传, 在指定范围内进行交互共享, 进行排序, 使得大家关注自身健康, 定期做检查 [0007] In order to solve the above problems, the present invention proposes a body network health information monitoring based on fuzzy neural network. Measuring and interacting system, the system uses fuzzy neural network algorithm to upload the user's health information in time, share and sort in the specified range, and make everyone pay attention to their health, and check regularly.
[0008] 为了实现上述目的, 本发明采用如下技术方案: [0008] In order to achieve the above object, the present invention adopts the following technical solutions:
[0009] 一种基于模糊神经网络的体域网健康信息监测、 交互系统, 包括采集端、 模糊 神经网络分析检测系统、 交互模块、 识别单元和显示单元; [0009] A body network health information monitoring and interaction system based on a fuzzy neural network, comprising a collecting end, a fuzzy neural network analyzing and detecting system, an interaction module, an identifying unit and a display unit;
[0010] 所述采集端包括采集模块、 控制模块, 所述采集模块用于采集家用医疗设备采 集的相关生物数据, 并发送给控制模块, 所述控制模块还用于控制采集模块进 行采集数据; [0010] The collection end includes an acquisition module and a control module, and the collection module is configured to collect related biological data collected by the home medical device, and send the data to the control module, where the control module is further configured to control the collection module to collect data;
[001 1 ] 所述交互模块, 用于建立交互群, 选择交互群大小、 相互进行数据交互的采集 端数目, 建立数据共享; [001 1] The interaction module is configured to establish an interaction group, select an interaction group size, a number of collection ends that perform data interaction with each other, and establish data sharing;
[0012] 所述模糊神经网络分析检测系统, 包括模型构建模块、 模糊神经网络数据分析 模块和排序模块, 其中: [0012] The fuzzy neural network analysis and detection system comprises a model building module, a fuzzy neural network data analysis module and a sorting module, wherein:
[0013] 所述模型构建模块, 接收交互模块中交互群内传输的生物数据, 并将其作为输 入变量, 通过输入层、 模糊化层模块, 构建生物数据的隶属度函数; [0013] the model construction module receives the biological data transmitted in the interaction group in the interaction module, and uses it as an input variable to construct a membership function of the biological data through the input layer and the fuzzy layer module;
[0014] 所述模糊神经网络数据分析模块, 用于通过构建模糊推理层、 归一层、 去模糊 层和输出层模块对生物数据进行处理, 计算输出生物指数值; [0014] the fuzzy neural network data analysis module is configured to process the biological data by constructing a fuzzy inference layer, a layer back layer, a deblurring layer, and an output layer module, and calculate an output biological index value;
[0015] 所述排序模块, 用于根据生物指数值, 在该交互群内进行数据排名; [0015] the sorting module is configured to perform data ranking in the interaction group according to the biometric index value;
[0016] 所述识别单元, 接收排序模块传输的数据, 读取交互模块的交互群, 识别同一 交互群内的采集模块、 显示单元, 将生物数据和数据排名通过显示单元进行显 示。 [0016] The identification unit receives the data transmitted by the sorting module, reads the interaction group of the interaction module, identifies the collection module and the display unit in the same interaction group, and displays the biological data and the data ranking through the display unit.
[0017] 所述采集模块, 实时监测通过家用医疗设备上传的数据, 包括收缩压、 舒张压 和心率。 [0017] The acquisition module monitors data uploaded by the home medical device in real time, including systolic blood pressure, diastolic blood pressure, and heart rate.
[0018] 所述模型构建模块, 包括输入层模块和模糊化层模块, 其中, 所述输入层模块 为第一层模块, 其输入值为家用医疗设备测量后上传的数据, 上传的数据包括 收缩压 SSY、 舒张压 SZY和心率 XL ; 仅起到传递的作用, 即将输入数据传递到下 一层模块, 输出值为: 0 SSY (1) =SSY, 0 SZY(1) =SZY, 0 XL(" =XL; 其中, 0 表示 第 X层模块的输出的 i值。
[0019] 所述模糊化层模块, 为第二层模块, 将输入数据划分模糊子集, 收缩压 SSY、 舒张 -压 SZY和心率 XL的模糊子集为 {正常 (M), 偏高 (H), 极高 (VH) }, 隶属度分别 为 SSY (M,H,VH)、 μ SZY (M, H, VH)和 XL (M,H,VH)。 [0018] The model building module includes an input layer module and a fuzzification layer module, wherein the input layer module is a first layer module, and the input value is data uploaded after measurement by the home medical device, and the uploaded data includes shrinkage. Pressure SSY, diastolic pressure SZY and heart rate XL; only play the role of transmission, the input data is passed to the next module, the output value is: 0 SSY (1 ) =SSY, 0 SZY (1) =SZY, 0 X L ("=XL; where 0 is the value of i of the output of the X-th layer module. [0019] The fuzzification layer module is a second layer module, and the input data is divided into a fuzzy subset, and the fuzzy subset of the systolic pressure SSY, the diastolic-pressure SZY, and the heart rate XL is {normal (M), high (H) ), very high (VH) }, membership degrees are SSY (M, H, VH), μ SZY (M, H, VH) and XL (M, H, VH).
[0020] 所述模糊化 2层模块, 对于 "正常 (M) " , 采用 Sigmoid函数的对称函数, 隶属 度为: [0020] The fuzzification layer 2 module, for "normal (M)", uses a symmetric function of the Sigmoid function, and the membership degree is:
[] []
=【1 + esp( (x― ,2 ] -1 =[1 + esp( (x― ,2 ] -1
[0021] 对于 "极鬲 (VH) " , 米用 Sigmoid函数, 隶属度为: [0021] For "polar 鬲 (VH)", the meter uses the Sigmoid function, and the membership is:
[] ^ -[! + es¾p《- ,Ι(χ— IF2¾j~I [] ^ -[! + es3⁄4p"- , Ι(χ— IF23⁄4j~ I
[0022] 对于 "偏高 (H) " , 采用高斯型函数, 隶属度为: [0022] For "high (H)", a Gaussian function is used, and the membership is:
[] []
[0023] Wl、 W2 为系统 , 初始值任意, 通过网络训练和自学习不断调整, 使网络实际 输出值与导师信号误差最小, 该层的节点输出范围在 0〜1 之间。 [0023] Wl, W2 are systems, the initial value is arbitrary, and the network training and self-learning are continuously adjusted to minimize the error between the actual output value of the network and the tutor signal. The node output range of this layer is between 0 and 1.
[0024] 对语言变量 M和 VH 的隶属度函数, 其中参数 W2 的作用是使隶属度函数沿水平 轴向右移; W1 的作用是调节隶属度函数的形状; 对语言变量 H的隶属度函数, W2和 W1分别表示隶属度函数的中心和宽度, 输入样本越靠近节点的中心, 输 出越大。 [0024] For the membership function of the linguistic variables M and VH, wherein the function of the parameter W2 is to shift the membership function to the right in the horizontal axis; the role of W1 is to adjust the shape of the membership function; the membership function of the linguistic variable H W2 and W1 represent the center and width of the membership function, respectively. The closer the input sample is to the center of the node, the larger the output.
[0025] 第二层模块的最终输出为: [0025] The final output of the second layer module is:
[0026] 式中, y=l,2,3, R取值为 SSY、 SZY、 XL, 分别对应于收缩压、 舒张压和心率 的隶属度函数。 In the formula, y=l, 2, 3, and R have values of SSY, SZY, and XL, which correspond to the membership functions of systolic blood pressure, diastolic blood pressure, and heart rate, respectively.
[0027] 所述模糊神经网络数据分析模块, 包括模糊推理层模块, 设为第三层模块, 用 于建立模糊推理规则库, 其中的语言规则是模糊逻辑推理的依据, 每条模糊推 理规则的结论参数为模糊子集{健康 (H), 亚健康 (SubH), 不健康 (UnH)}; 由于 输入数据 SSY、 SZY和 XL的模糊集具有 3个模糊标记, 因此模糊神经网络模型共有 推理规则 3X3X3=27条, 即为 27个输出节点; 对于任一组输入数据, 总可以找 到一条与之对应的推理规则; 每条推理规则的适用度 a( 1)为:
[0027] The fuzzy neural network data analysis module includes a fuzzy inference layer module, which is set as a third layer module, and is used to establish a fuzzy inference rule base, wherein the language rule is a basis of fuzzy logic reasoning, and each fuzzy inference rule is Conclusion The parameters are fuzzy subsets {health (H), sub-health (SubH), unhealthy (UnH)} ; since the fuzzy sets of input data SSY, SZY and XL have 3 fuzzy markers, the fuzzy neural network model has a common inference rule 3X3X3 = 27, which is 27 output nodes; for any set of input data, a corresponding inference rule can be found; the applicability a(1) of each inference rule is:
[0028] 式中, a( 1)表示第 1条推理规则的适用度, :1,2,3; j [0028] where a(1) represents the applicability of the first inference rule, :1,2,3; j
[0029] 所述模糊神经网络数据分析模块, 包括第四层模块, 即归一层模块, 节点总数 与第三层模块相同, 实现对第四层输出的归一化处理, 为解模糊做准备, 其输 出为: [0029] The fuzzy neural network data analysis module includes a fourth layer module, that is, a layer one module, and the total number of nodes is the same as that of the third layer module, realizing normalization processing on the fourth layer output, preparing for deblurring , its output is:
[0030] 其中 1=1,2— 27。 [0030] where 1=1, 2-27.
[0031] 所述模糊神经网络数据分析模块, 还包括去模糊层模块, 用于将经过模糊推理 得到的是对应于不同输出 (H,SubH,UnH) 的不同规则的隶属度, 通过去模糊的 方法得到三种健康状况的隶属度, 该层输出为: [0031] The fuzzy neural network data analysis module further includes a deblurring layer module, configured to obtain, by fuzzy inference, membership degrees of different rules corresponding to different outputs (H, SubH, UnH), by deblurring The method obtains the membership of three health conditions, and the output of this layer is:
[]
[0032] 其中, n表示该层输出节点的个数, n=l, 2, 3; 引入权值 W nl, 通过它在网络 中的变化调整各条模糊推理规则在判别的相对重要性。 [] [0032] wherein n represents the number of output nodes of the layer, n=l, 2, 3; the weight W nl is introduced, and the relative importance of each fuzzy inference rule is discriminated by the change in the network.
[0033] 权值 W nl的应用优化了模糊推理规则, 并使模糊推理具有自适应性, 从而提高 模糊神经网络模型的智能化程度。 [0033] The application of the weight W nl optimizes the fuzzy inference rule and makes the fuzzy reasoning adaptive, thereby improving the intelligence degree of the fuzzy neural network model.
[0034] 所述模糊神经网络数据分析模块, 还包括输出层模块, 为第六层模块, 用于利 用面积重心法, 通过权值 W i ( i=l,2,3 ) 调整三个输出模糊子集 (H,SubH,UnH) 的相对重要性, 取 W
即输出模糊子集中健康 (H) 对健康 指数值的贡献为 1, 亚健康 (SubH) 对健康指数值的贡献为 0. 5, 不健康 (UnH) 对健康指数值的贡献为 0; 其最终输出健康指数值 P为: [0034] The fuzzy neural network data analysis module further includes an output layer module, which is a sixth layer module, configured to adjust three output blurs by using a weight center method by using a weight W i (i=l, 2, 3) The relative importance of the subset (H, SubH, UnH), taking W That is, the output of the fuzzy subset (H) contributes to the health index value of 1, the sub-health (SubH) contributes to the health index value of 0.5, and the unhealthy (UnH) contributes to the health index value of 0; The health index value P is:
[0035] 所述模糊神经网络数据分析模块, 还包括 BP神经元网络学习训练模块, 用于通 过采用附加动量项改进的 BP 算法对网络进行训练, 利用误差反向传播的梯度 下降算法调整权值, 使权值按沿目标函数梯度变化的反方向进行调整, 确认模 糊神经网络中的参数有 Wl、 W2和 W nl, 先调整 W nl, 然后调整 Wl、 W2。 最终使 网络实际输出与期望输出的误差总均方差最小。 [0035] The fuzzy neural network data analysis module further includes a BP neural network learning training module, configured to train the network by using an improved BP algorithm with an additional momentum term, and adjust the weight by using a gradient descent algorithm of error back propagation. the weight in the objective function gradient along the reverse direction is adjusted to confirm the fuzzy neural network parameters Wl, W2 and W nl, first adjust W nl, then adjust Wl, W2. Ultimately, the total mean square error of the actual output of the network and the expected output is minimized.
[0036] 所述 BP神经元网络学习训练模块网络的误差信号为: e (n) = P e* [0036] The error signal BP neural network learning and training module of the network is: e (n) = P e *
-P (n)。 其中 P (n)为第 n次迭代的实际输出, P 为期望输出。 学习训练的目的即 最小化误差函数为
-P (n). Where P(n) is the actual output of the nth iteration and P is the desired output. The purpose of learning training is to minimize the error function.
[0037] 为了既能提高学习速度又能保持参数变化的稳定性, 采用基于广义 delta规则 的梯度下降法, 即
[0038] 其中,
[0037] In order to improve the learning speed and maintain the stability of the parameter change, a gradient descent method based on the generalized delta rule is adopted, that is, [0038] wherein
[0039] 式中, β为学习速率, 0〈β〈1, n AW n)为 Wherein β is the learning rate, 0<β<1, n AW n)
一般取值 0.9。 Generally, the value is 0.9.
[0040] 由此, 权值 W 的调整过程为: [0040] Thus, the adjustment process of the weight W is:
[0041] 求出 AW后, 可得修正后的权值 W nl, 即 W nl(n+l)= W nl(n) +AW [0041] After the AW is obtained, the corrected weight W nl can be obtained, that is, W nl (n+l)= W nl (n) +AW
[0042] 采用类似思路推导 Wl、 W2, [0042] Using a similar idea to derive Wl, W2,
[0043] 至此完成 Wl、 W2和 W nl的优化过程, Wl、 W2表现为隶属函数的自动生成, W nl 的变化表现为相应模糊规则相对重要性的改变, 通过训练最终提高了网络模糊 推理精度, 使实际输出更接近理想输出。 [0043] So far, the optimization process of Wl, W2 and W nl is completed, Wl and W2 are automatically generated as membership functions, and the change of W nl is represented by the change of the relative importance of the corresponding fuzzy rules, and the accuracy of network fuzzy inference is finally improved through training. , making the actual output closer to the ideal output.
[0044] 所述排序模块, 读取输出层模块的输出值, 读取该输出值对应的采集模块, 确 认该采集模块通过控制模块在交互模块中所选择的交互群, 采集交互群内其他 输出层模块的输出值, 在该交互群内进行数据排名。 [0044] the sorting module reads an output value of the output layer module, reads an acquisition module corresponding to the output value, and confirms that the collection module collects other outputs in the interaction group through the interaction group selected by the control module in the interaction module. The output value of the layer module, and the data is ranked within the interaction group.
发明的有益效果 Advantageous effects of the invention
有益效果 Beneficial effect
[0045] (1) 实现了健康测量数据的及时上传、 可选择范围内的交互, 在保护用户的 隐私、 尊重用户意愿的基础上进行信息的交互; [0045] (1) realizing timely uploading of health measurement data, interaction within a selectable range, and performing information interaction on the basis of protecting user privacy and respecting the user's will;
[0046] (2) 通过模糊逻辑和神经网络相结合的方式, 对体域网中医疗设备输出的各
种健康数据送入模糊神经网络系统模型进行分析, 得到健康指数值, 并进行健 康排名, 可以激发用户的积极性, 并让他们及时关注自身健康问题, 同时可以 按时使用家用医疗设备进行测量; [0046] (2) by means of a combination of fuzzy logic and a neural network, each of the medical device outputs in the body area network The health data is sent to the fuzzy neural network system model for analysis, and the health index value is obtained, and the health ranking is performed, which can stimulate the enthusiasm of the user, and let them pay attention to their own health problems in time, and can use the home medical equipment to measure on time;
[0047] ( 3 ) 系统具有很好的可扩展性, 可以加入更多其他类型的健康数据进行交互[0047] (3) The system has good scalability, and can join more other types of health data for interaction.
、 综合评测。 , comprehensive evaluation.
对附图的简要说明 Brief description of the drawing
附图说明 DRAWINGS
[0048] 图 1为本发明的结构示意图。 1 is a schematic structural view of the present invention.
[0049] 图 2为本发明的模糊神经网络模型图。 2 is a fuzzy neural network model diagram of the present invention.
[0050] 图 3为本发明的隶属度函数图。 3 is a graph of membership function of the present invention.
发明实施例 Invention embodiment
本发明的实施方式 Embodiments of the invention
[0051] 如图 1所示, 一种基于模糊神经网络的体域网健康信息监测、 交互系统, 包括 采集端、 模糊神经网络分析检测系统、 交互模块、 识别单元和显示单元; [0051] As shown in FIG. 1 , a body network health information monitoring and interaction system based on a fuzzy neural network includes a collection end, a fuzzy neural network analysis and detection system, an interaction module, an identification unit, and a display unit.
[0052] 所述采集端包括采集模块、 控制模块, 所述采集模块用于采集家用医疗设备采 集的相关生物数据, 并发送给控制模块, 所述控制模块还用于控制采集模块进 行采集数据; [0052] The collection end includes an acquisition module and a control module, and the collection module is configured to collect related biological data collected by the home medical device, and send the data to the control module, where the control module is further configured to control the collection module to collect data;
[0053] 所述交互模块, 用于建立交互群, 选择交互群大小、 相互进行数据交互的采集 端数目, 建立数据共享; [0053] the interaction module is configured to establish an interaction group, select an interaction group size, a number of collection ends that perform data interaction with each other, and establish data sharing;
[0054] 所述模糊神经网络分析检测系统, 包括模型构建模块、 模糊神经网络数据分析 模块和排序模块, 其中: [0054] The fuzzy neural network analysis and detection system comprises a model construction module, a fuzzy neural network data analysis module and a sorting module, wherein:
[0055] 所述模型构建模块, 接收交互模块中交互群内传输的生物数据, 并将其作为输 入变量, 通过输入层、 模糊化层模块, 构建生物数据的隶属度函数; [0055] the model building module receives the biological data transmitted in the interaction group in the interaction module, and uses it as an input variable to construct a membership function of the biological data through the input layer and the fuzzification layer module;
[0056] 所述模糊神经网络数据分析模块, 用于通过构建模糊推理层、 归一层、 去模糊 层和输出层模块对生物数据进行处理, 计算输出生物指数值; [0056] the fuzzy neural network data analysis module is configured to process the biological data by constructing a fuzzy inference layer, a layer back layer, a deblurring layer, and an output layer module, and calculate an output biological index value;
[0057] 所述排序模块, 用于根据生物指数值, 在该交互群内进行数据排名; [0057] the sorting module is configured to perform data ranking in the interaction group according to the biometric index value;
[0058] 所述识别单元, 接收排序模块传输的数据, 读取交互模块的交互群, 识别同一 交互群内的采集模块、 显示单元, 将生物数据和数据排名通过显示单元进行显
不。 [0058] the identification unit receives the data transmitted by the sorting module, reads the interaction group of the interaction module, identifies the collection module and the display unit in the same interaction group, and ranks the biological data and the data through the display unit. Do not.
[0059] 所述采集模块, 实时监测通过家用医疗设备上传的数据, 包括收缩压、 舒张压 和心率。 [0059] The collection module monitors data uploaded by the home medical device in real time, including systolic blood pressure, diastolic blood pressure, and heart rate.
[0060] 模糊神经网络模型包括输入层、 模糊化层、 模糊推理层、 归一化层、 去模糊层 和输出层, 如图 2所示。 对于每一层的输出变量, 有两个作用, 一个是用于训练 数据的反向输入, 一个是实际输出。 第三层为模糊推理层, 其每一节点代表一 条模糊规则。 因此, 所有的第三层节点组成了模糊规则库。 第二层和第三层之 间的连接作为连接推理机, 避免了模糊蕴涵推理过程。 第二层的连接定义了模 糊规则的前件, 第三层的连接定义了模糊规则的后件。 正常传播方向为从第一 层到第六层, 即从下向上, 在学习训练过程中, 信号是从第六层到第一层传播 的, 即从上向下。 各层的输入输出关系如下 (0(x)表示第 X层的输出) 。 [0060] The fuzzy neural network model includes an input layer, a fuzzification layer, a fuzzy inference layer, a normalization layer, a deblurring layer, and an output layer, as shown in FIG. 2. For each layer of output variables, there are two functions, one for the inverse input of the training data and one for the actual output. The third layer is the fuzzy inference layer, and each node represents a fuzzy rule. Therefore, all Layer 3 nodes constitute a fuzzy rule base. The connection between the second layer and the third layer acts as a connection inference engine, avoiding the fuzzy implication reasoning process. The second layer of the connection defines the predecessor of the fuzzy rule, and the third layer of the connection defines the posterior of the fuzzy rule. The normal direction of propagation is from the first layer to the sixth layer, that is, from bottom to top. During the learning and training process, the signal propagates from the sixth layer to the first layer, that is, from top to bottom. The input and output relationship of each layer is as follows (0 (x) indicates the output of the Xth layer).
[0061] (1) 第一层为输入层。 输入值为家用医疗设备测量后上传的数据, 上传的数 据包括收缩压 SSY、 舒张压 SZY和心率 XL。 该层仅起到传递的作用, 即将输入数 据传递到下一层, 输出为: [0061] (1) The first layer is an input layer. The input value is the data uploaded after measurement by the home medical device. The uploaded data includes systolic blood pressure SSY, diastolic blood pressure SZY and heart rate XL. This layer only serves as a pass, passing the input data to the next layer. The output is:
[0063] (2) 第二层为模糊化层。 该层将输入数据划分模糊子集, 收缩压 SSY、 舒张压 SZY和心率 XL的模糊子集为 {正常 (M), 偏高 (H), 极高 (VH)}, 隶属度分别为 WSSi (M,H,VH)、 SZY(M,H,VH)和 μΧί(Μ,Η,νΗ)。 (本专利实施案例主要针对高血压 人群的健康状况, 这里将低血压、 低心率归于正常范畴) [0063] (2) The second layer is a fuzzification layer. This layer divides the input data into fuzzy subsets. The fuzzy subsets of systolic pressure SSY, diastolic pressure SZY and heart rate XL are {normal (M), high (H), very high (VH)}, and the membership is W SSi. (M, H, VH), SZY (M, H, VH) and μ Χί (Μ, Η, νΗ). (The implementation case of this patent is mainly for the health status of high blood pressure people, where the low blood pressure and low heart rate are attributed to the normal range)
[0064] 对于 "正常 (Μ) ", 采用 Sigmoid函数的对称函数, 隶属度为: [0064] For "normal (Μ)", a symmetric function using the Sigmoid function, the membership is:
[] []
= [l +€¾ (Ι: ( - ,2朋 _Ι = [L + € ¾ (Ι : (-, 2 Peng _Ι
[0065] 对于 "极高 (VH) ", 采用 Sigmoid函数, 隶属度为: [0065] For "very high (VH)", the Sigmoid function is used, and the membership is:
[] []
[0066] 对于 "偏高 (H) ", 采用高斯型函数, 隶属度为:
[0067] Wl、 W2 初始值任意, 通过网络训练和自学习不断调整, 使网络实际输出值与 a [0066] For "higher (H)", a Gaussian function is used, and the membership is: [0067] Wl, W2 initial value is arbitrary, continuously adjusted through network training and self-learning, so that the actual output value of the network and a
导师信号误差最小, 经调整后的权值即可用于实际的健康状况评测。 节点输出 范围在 0 之间。 对语言变量 M和 VH 的隶属度函数, 其中参数 W2 的作用是使 隶属度函数沿水平轴向右移; W1 的作用是调节隶属度函数的形状, 较大的 W1 使函数逼近阶跃, 较 W1 使函数变得较为平坦。 对语言变量 H的隶属度函数 , W2 和 W1 分别表示隶属度函数的中心和宽度, 输入样本越靠近节点的中心, 输出越大。 The tutor's signal error is minimal, and the adjusted weight can be used for actual health assessment. The node output range is between 0. For the membership function of the linguistic variables M and VH, the role of the parameter W2 is to shift the membership function to the right in the horizontal axis; the role of W1 is to adjust the shape of the membership function, and the larger W1 makes the function approximate the step. W1 makes the function flatter. For the membership function of the linguistic variable H, W2 and W1 represent the center and width of the membership function, respectively. The closer the input sample is to the center of the node, the larger the output.
[0068] 第二层的最终输出为: [0068] The final output of the second layer is:
[0069] 式中, y=l,2,3, R取值为 SSY、 SZY、 XL, 分别对应于收缩压、 舒张压和心率的 隶属度函数。 In the formula, y=l, 2, 3, and R have values of SSY, SZY, and XL, which correspond to the membership functions of systolic blood pressure, diastolic blood pressure, and heart rate, respectively.
[0070] ( 3) 第三层为模糊推理层。 模糊推理中最重要的过程是建立模糊推理规则库 , 其中的语言规则是模糊逻辑推理的依据, 每条模糊推理规则的结论参数为模 糊子集 {健康 (H), 亚健康(SubH), 不健康 (UnH) }。 由于输入数据 SSY、 SZY和 XL 的模糊集具有 3个模糊标记, 因此模糊神经网络模型共有推理规则 3 X 3 X 3=27条 。 每条推理规则的适用度 a (l)为: [0070] (3) The third layer is a fuzzy inference layer. The most important process in fuzzy reasoning is to establish a fuzzy inference rule base, in which the linguistic rules are the basis of fuzzy logic reasoning, and the conclusion parameters of each fuzzy inference rule are fuzzy subsets {health (H), sub-health (SubH), unhealthy (UnH) }. Since the fuzzy sets of the input data SSY, SZY and XL have three fuzzy marks, the fuzzy neural network model has a common inference rule of 3 X 3 X 3=27 . The applicability of each inference rule a (l) is:
[0071] 式中, a (l)表示第 1条推理规则的适用度, 1=1, 2— 27 ; i=l,2,3 ; j=l,2,3 ;
表 1所示, 模糊规则用语言变量表述为: In the formula, a (l) represents the applicability of the first inference rule, 1=1, 2-27; i=l, 2, 3; j=l, 2, 3; As shown in Table 1, fuzzy rules are expressed in terms of linguistic variables:
[0073] Rulel : IF SSY is M and SZY is M and XL is M then C is H. [0073] Rulel: IF SSY is M and SZY is M and XL is M then C is H.
[0074] Rule2: IF SSY is M and SZY is H and XL is M then C is H. [0074] Rule2 : IF SSY is M and SZY is H and XL is M then C is H.
[0075] [0075]
[0076] Rule26: IF SSY is VH and SZY is H and XL is VH then C is UnH. [0076] Rule26: IF SSY is VH and SZY is H and XL is VH then C is UnH.
[0077] Rule27: IF SSY is VH and SZY is VH and XL is VH then C is UnH. [0077] Rule27: IF SSY is VH and SZY is VH and XL is VH then C is UnH.
[0078] 其中 C为处于某种健康状况的概率。 Where C is the probability of being in a certain state of health.
[0079] 表 1 Table 1
[]
[]
[表 1] [Table 1]
[0081] ( 4) 第四层为归一化层。 节点总数与第三层相同, 实现对第四层输出的归一 化处理, 为解模糊做准备, 其输出为: [0081] (4) The fourth layer is a normalized layer. The total number of nodes is the same as that of the third layer, and the normalized processing of the fourth layer output is implemented to prepare for the deblurring. The output is:
[0082] 其中 1=1, 2···27 ο Where 1=1, 2···27 ο
[0083] ( 5 ) 第五层为去模糊层。 经过模糊推理得到的是对应于不同输出 (H,SubH,Un H) 的不同规则的隶属度, 还需要通过去模糊的方法得到三种健康状况的隶属度 , 该层输出为:
[0083] (5) The fifth layer is a deblurring layer. The fuzzy inference results in the membership of different rules corresponding to different outputs (H, SubH, Un H), and the degree of membership of the three health conditions needs to be obtained by deblurring. The output of this layer is:
[0084] 其中, n表示该层输出节点的个数, n=l,2,3。 引入权值 W nl, 通过它在网络中 的变化调整各条模糊推理规则在判别的相对重要性。 权值 W nl的应用优化了模糊 推理规则, 并使模糊推理具有自适应性, 从而提高模糊神经网络模型的智能化 [0084] wherein n represents the number of output nodes of the layer, n=l, 2, 3. The weight W nl is introduced, and the relative importance of each fuzzy inference rule is discriminated by its change in the network. The application of weight W nl optimizes the fuzzy inference rules and makes the fuzzy reasoning adaptive, thus improving the intelligence of the fuzzy neural network model.
[0085] ( 6 ) 第六层为输出层。 利用面积重心法, 通过权值 W i ( i=l,2,3 ) 调整三个 输出模糊子集 (H, SubH, UnH) 的相对重要性, 取 W W 2=0. 5, W 3=0, 即输出
模糊子集中健康 (H) 对健康指数值的贡献为 1, 亚健康 (SubH) 对健康指数值 的贡献为 0. 5, 不健康 (UnH) 对健康指数值的贡献为 0。 其最终输出健康指数值 P为: [0085] (6) The sixth layer is an output layer. Using the area center of gravity method, the relative importance of the three output fuzzy subsets (H, SubH, UnH) is adjusted by the weight W i ( i=l, 2, 3 ), taking WW 2 =0. 5, W 3 =0 Output The contribution in the fuzzy subset (H) to the health index value is 1, and the contribution of sub-health (SubH) to the health index value is 0.5, and the unhealthy (UnH) contribution to the health index value is zero. Its final output health index value P is:
[0086] 优选的, 还包括 BP神经元网络的学习训练过程。 模糊神经网络中需要确定的参 数有 Wl、 W2和 W nl。 采用附加动量项改进的 BP 算法对网络进行训练。 根据规 则学习, 利用误差反向传播的梯度下降算法调整权值, 使权值按沿目标函数梯 度变化的反方向进行调整, 因此, 先调整 W nl, 然后调整 Wl、 W2。 最终使网络 实际输出与期望输出的误差总均方差最小。 具体的网络学习算法运算过程如下 [0086] Preferably, a learning training process of the BP neural network is also included. The parameters that need to be determined in the fuzzy neural network are Wl, W2, and W nl . The network is trained using an improved BP algorithm with additional momentum terms. According to the rule learning, the gradient descent algorithm of error back propagation is used to adjust the weight, so that the weight is adjusted in the opposite direction of the gradient of the objective function. Therefore, W nl is adjusted first, then Wl and W2 are adjusted. Ultimately, the total mean square error of the actual output of the network and the expected output is minimized. The specific network learning algorithm operation process is as follows
[0087] 网络的误差信号为: e (n) = P e* _P (n)。 其中 P (n)为第 n次迭代的实际输出, P [0087] The error signal of the network is: e (n) = P e * _P (n). Where P (n) is the actual output of the nth iteration, P
为期望输出。 学习训练的目的即最小化误差函数为
For the desired output. The purpose of learning training is to minimize the error function.
。 为了既能提高学习速度又能保持参数变化的稳定性, 采用基于广义 delta规 则的梯度下降法, 即。
. In order to improve the learning speed and maintain the stability of the parameter change, a gradient descent method based on the generalized delta rule is adopted.
, 式中 β为学习速率, ο〈β〈1, n A W Where β is the learning rate, ο<β<1, n A W
(n)为动 项, n为动量系数, ο〈η〈ι, 一般取值 0. 9。
[0088] 由此, 权值 W _^的调整过程为: (n) is a dynamic term, n is a momentum coefficient, ο <η <ι, generally takes a value of 0.9. [0088] Thus, the adjustment process of the weight W _^ is:
[] []
Δ (η) = - ^ + (-) = - β ^ + 》 Δ (η) = - ^ + (-) = - β ^ + 》
[] ,, [] ,,
, * _ ρ、 ν / ui , * _ ρ, ν / u i
= β ^— ¾ ) + 扁 ― « = Β ^ - ¾) + flat - «
[0089] 求出 AW后, 可得修正后的权值 W nl, 即 W nl(n+l)= W nl(n+l)+ AW 1o 采用类 似思路推导 Wl、 W2o [0089] After the AW is obtained, the modified weight W nl can be obtained, that is, W nl (n+l)= W nl (n+l)+ AW 1o, and a similar idea is used to derive Wl, W2o.
ΔΙΤί(ϊΐ) = -Ρ— - t|A i( ) ΔΙΤί(ϊΐ) = -Ρ— - t|A i( )
[0090] 至此完成 Wl、 W2和 W nl的优化过程, Wl、 W2表现为隶属函数的自动生成, W n] 的变化表现为相应模糊规则相对重要性的改变, 通过训练最终提高了网络模糊 推理精度, 使实际输出更接近理想输出。 [0090] This completes Wl, W2 and W nl optimization process, and Wl, W2 performance to automatically generate membership functions, W n] changes the performance of the corresponding changes in the relative importance of fuzzy rules, through training and ultimately improve the network fuzzy reasoning Precision, making the actual output closer to the ideal output.
[0091] 本发明的其他实施案例中输入变量还可以为血氧值、 血糖值等。 [0091] In other embodiments of the invention, the input variables may also be blood oxygenation values, blood glucose values, and the like.
[0092] 所述排序模块, 读取输出层模块的输出值, 读取该输出值对应的采集模块, 确 认该采集模块通过控制模块在交互模块中所选择的交互群, 采集交互群内其他 输出层模块的输出值, 在该交互群内进行数据排名。 [0092] the sorting module reads an output value of the output layer module, reads an acquisition module corresponding to the output value, and confirms that the collection module collects other outputs in the interaction group through the interaction group selected by the control module in the interaction module. The output value of the layer module, and the data is ranked within the interaction group.
[0093] 上述虽然结合附图对本发明的具体实施方式进行了描述, 但并非对本发明保护 范围的限制, 所属领域技术人员应该明白, 在本发明的技术方案的基础上, 本 领域技术人员不需要付出创造性劳动即可做出的各种修改或变形仍在本发明的 保护范围以内。
The above description of the specific embodiments of the present invention has been described with reference to the accompanying drawings, but is not intended to limit the scope of the present invention. It should be understood by those skilled in the art that Various modifications or variations that can be made by the creative work are still within the scope of the invention.
Claims
权利要求书 Claim
[权利要求 1 ] 一种基于模糊神经网络的体域网健康信息监测、 交互系统, 其特征是 [Claim 1] A fuzzy neural network based body area network health information monitoring and interaction system, characterized in that
: 包括采集端、 模糊神经网络分析检测系统、 交互模块、 识别单元和 显示单元; : including an acquisition end, a fuzzy neural network analysis detection system, an interaction module, an identification unit, and a display unit;
所述采集端包括采集模块、 控制模块, 所述采集模块用于采集家用医 疗设备采集的相关生物数据, 并发送给控制模块, 所述控制模块还用 于控制采集模块进行采集数据; The collection end includes an acquisition module and a control module, and the collection module is configured to collect related biological data collected by the home medical device, and send the data to the control module, where the control module is further configured to control the collection module to collect data;
所述交互模块, 用于建立交互群, 选择交互群大小、 相互进行数据交 互的采集端数目, 建立数据共享; The interaction module is configured to establish an interaction group, select an interaction group size, and the number of collection ends that perform data interaction with each other to establish data sharing;
所述模糊神经网络分析检测系统, 包括模型构建模块、 模糊神经网络 数据分析模块和排序模块, 其中: The fuzzy neural network analysis and detection system comprises a model building module, a fuzzy neural network data analysis module and a sorting module, wherein:
所述模型构建模块, 接收交互模块中交互群内传输的生物数据, 并将 其作为输入变量, 通过输入层、 模糊化层模块, 构建生物数据的隶属 度函数; The model building module receives the biological data transmitted in the interaction group in the interaction module, and uses it as an input variable to construct a membership function of the biological data through the input layer and the fuzzy layer module;
所述模糊神经网络数据分析模块, 用于通过构建模糊推理层、 归一层 、 去模糊层和输出层模块对生物数据进行处理, 计算输出生物指数值 所述排序模块, 用于根据生物指数值, 在该交互群内进行数据排名; 所述识别单元, 接收排序模块传输的数据, 读取交互模块的交互群, 识别同一交互群内的采集模块、 显示单元, 将生物数据和数据排名通 过显示单元进行显示。 The fuzzy neural network data analysis module is configured to process the biological data by constructing a fuzzy inference layer, a layer back layer, a deblurring layer and an output layer module, and calculate an output biometric value value sorting module, which is used according to the biological index value And performing data ranking in the interaction group; the identification unit receives data transmitted by the sorting module, reads an interaction group of the interaction module, identifies an acquisition module and a display unit in the same interaction group, and displays the biological data and the data by displaying The unit is displayed.
[权利要求 2] 如权利要求 1所述的系统, 其特征是: 所述采集模块, 实时监测通过 家用医疗设备上传的数据, 包括收缩压、 舒张压和心率。 [Claim 2] The system of claim 1, wherein: the acquisition module monitors data uploaded by the home medical device in real time, including systolic blood pressure, diastolic blood pressure, and heart rate.
[权利要求 3] 如权利要求 1所述的系统, 其特征是: 所述模型构建模块, 包括输入 层模块和模糊化层模块, 其中, 所述输入层模块为第一层模块, 其输 入值为家用医疗设备测量后上传的数据, 上传的数据包括收缩压 SSY 、 舒张压 SZY和心率 XL ; 仅起到传递的作用, 即将输入数据传递到下 一层模块, 输出值为: 0 SSY(" =SSY, 0 SZY(1) =SZY, 0 XL(» =XL ; 其中,
0 表示第 x层模块的输出的 i值。 [Claim 3] The system according to claim 1, wherein: the model building module comprises an input layer module and an obfuscation layer module, wherein the input layer module is a first layer module, and an input value thereof For the data uploaded after measurement of the home medical device, the uploaded data includes systolic pressure SSY, diastolic pressure SZY and heart rate XL; only for the transfer function, the input data is transferred to the next module, the output value is: 0 SSY (" =SSY, 0 SZY (1) =SZY, 0 XL (» =XL ; where, 0 represents the i value of the output of the xth layer module.
[权利要求 4] 如权利要求 3所述的系统, 其特征是: 所述模糊化层模块, 为第二层 模块, 将输入数据划分模糊子集, 收缩压 SSY、 舒张压 SZY和心率 XL的 模糊子集为 {正常 (M), 偏高 (H), 极高 (VH) }, 隶属度分别为 w SSY [Claim 4] The system according to claim 3, wherein: the fuzzification layer module is a second layer module that divides input data into a fuzzy subset, systolic pressure SSY, diastolic pressure SZY, and heart rate XL The fuzzy subset is {normal (M), high (H), extremely high (VH) }, and the membership is w SSY
(M,H,VH)、 μ SZY (M, H, VH)禾卩 XL (M, H,VH)。 (M, H, VH), μ SZY (M, H, VH) and XL (M, H, VH).
[权利要求 5] 如权利要求 4所述的系统, 其特征是: 所述模糊化层模块, 对于 "正 常 (M) " , 采用 Sigmoid函数的对称函数, 隶属度为: = [i + e¾ ( - 糊-1 对于 "极高 (VH) " , 采用 Sigmoid函数, 隶属度为: [Claim 5] The system according to claim 4, wherein: the fuzzification layer module, for "normal (M)", adopts a symmetric function of a Sigmoid function, and the membership degree is: = [i + e3⁄4 ( - Paste - 1 For "Extremely High (VH)", using the Sigmoid function, the membership is:
/ - [l + ejEp(-iri( - lf 2¾ ]_i 对于 "偏高 (H) " , 采用高斯型函数, 隶属度为:
/ - [l + ejEp(-iri( - lf 23⁄4 ] _i for "high (H)", using a Gaussian function, membership is:
Wl、 W2为系统, 初始值任意, 通过网络训练和自学习不断调整, 使 网络实际输出值与导师信号误差最小, 该层的节点输出范围在 0〜1 之间; Wl, W2 are the system, the initial value is arbitrary, and the network training and self-learning are continuously adjusted to minimize the error between the actual output value of the network and the tutor signal. The node output range of this layer is between 0~1;
对语言变量 M和 VH 的隶属度函数, 其中参数 W2 的作用是使隶属度函 数沿水平轴向右移; W1 的作用是调节隶属度函数的形状; 对语言变 量 H的隶属度函数, W2 和 W1 分别表示隶属度函数的中心和宽度, 输入样本越靠近节点的中心, 输出越大; The membership function of the linguistic variables M and VH, where the role of the parameter W2 is to shift the membership function to the right in the horizontal axis; the role of W1 is to adjust the shape of the membership function; the membership function of the linguistic variable H, W2 and W1 represents the center and width of the membership function, respectively. The closer the input sample is to the center of the node, the larger the output;
式中, y=l,2,3, R取值为 SSY、 SZY、 XL, 分别对应于收缩压、 舒张 压和心率的隶属度函数。 In the formula, y=l, 2, 3, and R are SSY, SZY, and XL, which correspond to the membership functions of systolic blood pressure, diastolic blood pressure, and heart rate, respectively.
[权利要求 6] 如权利要求 1所述的系统, 其特征是: [Clave 6] The system of claim 1 wherein:
块, 包括模糊推理层模块, 设为第三层模块, 用于建立模糊推理规则 库, 其中的语言规则是模糊逻辑推理的依据, 每条模糊推理规则的结 论参数为模糊子集 {健康 (H), 亚健康 (SubH), 不健康 (UnH)}; 由于输 入数据 SSY、 SZY和 XL的模糊集具有 3个模糊标记, 因此模糊神经网络 模型共有推理规则 3X3X3=27条, 即为 27个输出节点; 对于任一组输 入数据, 总可以找到一条与之对应的推理规则; 每条推理规则的适用 度 a( 1)为: : 2 % 式中, a( 1)表示第 1条推理规则的适用度, 1=1, 2— 27; i=l,2,3; J=l 2, 3; k=l, 2, 3; 第三层的输出为:
The block, including the fuzzy inference layer module, is set as a third layer module for establishing a fuzzy inference rule base, wherein the language rule is the basis of fuzzy logic reasoning, and the conclusion parameter of each fuzzy inference rule is a fuzzy subset {health (H Sub-health (SubH), unhealthy (UnH)} ; since the fuzzy sets of the input data SSY, SZY and XL have 3 fuzzy marks, the fuzzy neural network model has a total inference rule of 3X3X3=27, which is 27 output nodes. For any set of input data, a corresponding inference rule can be found; the applicability a(1) of each inference rule is: : 2 % where a(1) indicates the application of the first inference rule Degree, 1=1, 2-27; i=l,2,3; J=l 2, 3; k=l, 2, 3; The output of the third layer is:
[权利要求 7] 如权利要求 1所述的系统, 其特征是: 所述模糊神经网络数据分析模 块, 包括第四层模块, 即归一层模块, 节点总数与第三层模块相同, 实现对第四层输出的归一化处理, 为解模糊做准备, 其输出为:
[Claim 7] The system according to claim 1, wherein: the fuzzy neural network data analysis module comprises a fourth layer module, that is, a layer-by-layer module, and the total number of nodes is the same as that of the third layer module, and the pair is implemented. The normalization of the fourth layer of output, in preparation for deblurring, the output is:
1=1 l=! 1=1 l=!
其中 1=1, 2— 27; Where 1=1, 2-27;
所述模糊神经网络数据分析模块, 还包括去模糊层模块, 用于将经过 模糊推理得到的是对应于不同输出 (H,SubH,UnH) 的不同规则的隶属 度, 通过去模糊的方法得到三种健康状况的隶属度, 该层输出为:
其中, n表示该层输出节点的个数, n=l,2,3; 引入权值 W nl, 通过 它在网络中的变化调整各条模糊推理规则在判别的相对重要性。 The fuzzy neural network data analysis module further includes a deblurring layer module, which is used to obtain fuzzy membership in which the membership degrees of different rules corresponding to different outputs (H, SubH, UnH) are obtained by deblurring. The degree of membership of a health condition, the output of this layer is: Where n is the number of output nodes of the layer, n=l, 2, 3; the weight W nl is introduced, and the relative importance of each fuzzy inference rule is determined by its variation in the network.
[权利要求 8] 如权利要求 1所述的系统, 其特征是: 所述模糊神经网络数据分析模 块, 还包括输出层模块, 为第六层模块, 用于利用面积重心法, 通过 权值 W i ( i=l,2,3) 调整三个输出模糊子集 (H,SubH,UnH) 的相对重 要性, 取 W W 2=0. 5, W =0, 即输出模糊子集中健康 (H) 对健 康指数值的贡献为 1, 亚健康 (SubH) 对健康指数值的贡献为 0. 5, 不 健康 (UnH) 对健康指数值的贡献为 0; 其最终输出健康指数值 P为: [Claim 8] The system according to claim 1, wherein: the fuzzy neural network data analysis module further comprises an output layer module, which is a sixth layer module, configured to use an area center of gravity method, and pass the weight W i ( i=l,2,3) adjusts the relative importance of the three output fuzzy subsets (H, SubH, UnH), taking WW 2 =0. 5, W =0, ie output fuzzy subset health (H) The contribution to the health index value is 1, the sub-health (SubH) contribution to the health index value is 0.5, the unhealthy (UnH) contribution to the health index value is 0; and its final output health index value P is:
[权利要求 9] 如权利要求 1所述的系统, 其特征是: 所述模糊神经网络数据分析模 块, 还包括 BP神经元网络学习训练模块, 用于通过采用附加动量项改 进的 BP 算法对网络进行训练, 利用误差反向传播的梯度下降算法调
整权值, 使权值按沿目标函数梯度变化的反方向进行调整, 确认模糊 神经网络中的参数有 Wl、 W2和 W nl, 先调整 W nl, 然后调整 Wl、 W2 , 最终使网络实际输出与期望输出的误差总均方差最小。
[Claim 9] The system according to claim 1, wherein: the fuzzy neural network data analysis module further comprises a BP neural network learning training module, configured to improve the network by using an additional momentum item BP algorithm Training, using the gradient descent algorithm of error back propagation The weighting value is adjusted according to the direction opposite to the gradient of the objective function. It is confirmed that the parameters in the fuzzy neural network are Wl, W2 and W nl , W nl is adjusted first, then Wl and W2 are adjusted, and finally the actual output of the network is made. The total mean square error of the error with the desired output is minimal.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410756667.4A CN104484562B (en) | 2014-12-10 | 2014-12-10 | A kind of body area network health and fitness information monitoring based on fuzzy neural network, interactive system |
CN201410756667.4 | 2014-12-10 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016091102A1 true WO2016091102A1 (en) | 2016-06-16 |
Family
ID=52759103
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2015/096198 WO2016091102A1 (en) | 2014-12-10 | 2015-12-02 | Fuzzy neural network-based body area network health information monitoring and interaction system |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN104484562B (en) |
WO (1) | WO2016091102A1 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110473591A (en) * | 2019-08-20 | 2019-11-19 | 西南林业大学 | Idiotype network functional module based on quantum calculation is excavated and analysis method |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104484562B (en) * | 2014-12-10 | 2018-03-16 | 山东中弘信息科技有限公司 | A kind of body area network health and fitness information monitoring based on fuzzy neural network, interactive system |
CN109539886B (en) * | 2018-03-25 | 2021-06-01 | 哈尔滨工程大学 | Submarine stealth aid decision-making system and submarine stealth aid decision-making method |
CN110263620B (en) * | 2019-05-06 | 2021-06-01 | 杭州电子科技大学 | Based on L2,1Age estimation method for bias label learning |
CN110993085A (en) * | 2019-11-26 | 2020-04-10 | 浙江工业大学 | Blood sugar/blood pressure fuzzy monitoring method with privacy protection |
CN115394444A (en) * | 2022-09-06 | 2022-11-25 | 广东技术师范大学 | Health condition hierarchical quantitative evaluation implementation method and system based on fuzzy theory |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102811242A (en) * | 2011-08-05 | 2012-12-05 | 武汉荣坛技术有限公司 | Mobile health monitoring and service network system |
WO2014043894A1 (en) * | 2012-09-21 | 2014-03-27 | Nokia Corporation | Method and apparatus for providing access control to shared data based on trust level |
CN103891253A (en) * | 2013-03-08 | 2014-06-25 | 北京海银创业科技孵化器投资中心(有限合伙) | Showing method, system and device for social network carrier |
CN103976715A (en) * | 2014-06-09 | 2014-08-13 | 江苏启润科技有限公司 | Multifunctional human health self-examination system |
CN104484562A (en) * | 2014-12-10 | 2015-04-01 | 山东中弘信息科技有限公司 | Fuzzy neural network-based body area network health information monitoring and interacting system |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101485569B (en) * | 2008-09-26 | 2010-06-02 | 厦门大学 | Traditional Chinese medicine multifunctional intelligent diagnostic apparatus based on self-adapting fuzzy logic |
CN101739501A (en) * | 2008-11-04 | 2010-06-16 | 财团法人资讯工业策进会 | Identification system and method for risk degree of sickees |
CN102360405B (en) * | 2011-08-25 | 2015-08-05 | Tcl集团股份有限公司 | A kind of tele-medicine health monitoring data submit sort method and system to |
CN102799794B (en) * | 2012-08-28 | 2015-09-23 | 史荣涛 | The self-service evaluating system of life entity physiological situation |
CN103514357B (en) * | 2012-12-31 | 2017-07-11 | Tcl集团股份有限公司 | Remote health monitoring method and its monitor system |
CN103345152B (en) * | 2013-06-05 | 2016-04-20 | 重庆科技学院 | The healthy control method of a kind of domestic environment based on fuzzy intelligence Behavior modeling |
CN104158914A (en) * | 2014-09-10 | 2014-11-19 | 航天神舟生物科技集团有限公司 | Family-oriented remote urine-detecting and health-counseling service system and method |
-
2014
- 2014-12-10 CN CN201410756667.4A patent/CN104484562B/en active Active
-
2015
- 2015-12-02 WO PCT/CN2015/096198 patent/WO2016091102A1/en active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102811242A (en) * | 2011-08-05 | 2012-12-05 | 武汉荣坛技术有限公司 | Mobile health monitoring and service network system |
WO2014043894A1 (en) * | 2012-09-21 | 2014-03-27 | Nokia Corporation | Method and apparatus for providing access control to shared data based on trust level |
CN103891253A (en) * | 2013-03-08 | 2014-06-25 | 北京海银创业科技孵化器投资中心(有限合伙) | Showing method, system and device for social network carrier |
CN103976715A (en) * | 2014-06-09 | 2014-08-13 | 江苏启润科技有限公司 | Multifunctional human health self-examination system |
CN104484562A (en) * | 2014-12-10 | 2015-04-01 | 山东中弘信息科技有限公司 | Fuzzy neural network-based body area network health information monitoring and interacting system |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110473591A (en) * | 2019-08-20 | 2019-11-19 | 西南林业大学 | Idiotype network functional module based on quantum calculation is excavated and analysis method |
CN110473591B (en) * | 2019-08-20 | 2022-09-27 | 西南林业大学 | Gene network function module mining and analyzing method based on quantum computing |
Also Published As
Publication number | Publication date |
---|---|
CN104484562B (en) | 2018-03-16 |
CN104484562A (en) | 2015-04-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2016091102A1 (en) | Fuzzy neural network-based body area network health information monitoring and interaction system | |
Tan et al. | Non-invasive continuous blood pressure measurement based on mean impact value method, BP neural network, and genetic algorithm | |
Hernaez | Reliability and agreement studies: a guide for clinical investigators | |
Meraj et al. | Artificial intelligence in diagnosing tuberculosis: a review | |
CN105550509B (en) | A kind of fast evaluation method of medical infusion drop and system | |
JP6088139B2 (en) | Performing subject measurements | |
JP2018142259A (en) | Manufacturing management device, method, and program | |
Bourke et al. | Gait characteristics harvested during a smartphone-based self-administered 2-minute walk test in people with multiple sclerosis: test-retest reliability and minimum detectable change | |
Büyüközkan et al. | Smart medical device selection based on interval valued intuitionistic fuzzy VIKOR | |
JP7266548B2 (en) | Apparatus, program and method for estimating physical condition using training stress information | |
Rusnock et al. | Objective-analytical measures of workload–the third pillar of workload triangulation? | |
Devi et al. | Design methodology of a fuzzy knowledgebase system to predict the risk of diabetic nephropathy | |
Pradana et al. | Fuzzy Tsukamoto Implementation to Detect Physiological Condition on IoT-Based e-Learning Users | |
Zhi et al. | Design of intelligent rehabilitation evaluation scale for stroke patients based on genetic algorithm and extreme learning machine | |
Ngafidin et al. | Diagnosis of Tuberculosis by Using a Fuzzy Logic Expert System | |
CN116019429A (en) | Health monitoring method, device, equipment and storage medium based on physiological index | |
Jung et al. | User Pattern Learning Algorithm based MDSS (Medical Decision Support System) Framework under Ubiquitous | |
Selvan et al. | Healthcare Application System with Cyber-Security Using Machine Learning Techniques | |
CN105760702B (en) | A kind of general practice system with monitoring camera-shooting | |
Zakynthinaki et al. | A Neural Network Model for Estimating the Heart Rate Response to Constant Intensity Exercises | |
Subhedar et al. | A real time healthcare monitoring system based on open source IoT and ANFIS | |
Muthukumar et al. | A mean potentiality approach of an intuitionistic fuzzy soft set based decision making problem in medical science | |
TWI529652B (en) | Influenza-like illness diagnostic assistance system | |
CN116204793B (en) | Industrial Internet cloud edge model aggregation method based on federal learning | |
Khorozov et al. | Vital signs monitoring using fuzzy logic rules |
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: 15867793 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: 15867793 Country of ref document: EP Kind code of ref document: A1 |