CN108597609A - A kind of doctor based on LSTM networks is foster to combine health monitor method - Google Patents
A kind of doctor based on LSTM networks is foster to combine health monitor method Download PDFInfo
- Publication number
- CN108597609A CN108597609A CN201810417954.0A CN201810417954A CN108597609A CN 108597609 A CN108597609 A CN 108597609A CN 201810417954 A CN201810417954 A CN 201810417954A CN 108597609 A CN108597609 A CN 108597609A
- Authority
- CN
- China
- Prior art keywords
- data
- health
- model
- years old
- training
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
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
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
It is supported the invention discloses a kind of doctor based on LSTM networks and combines health monitor method, included the following steps：Via health detection equipment, smart home and HIS systems, acquisition is collected 60 years old and the health and fitness information data in all directions of the above crowd, and be stored in cloud database；Denoising is carried out using the adaptive notch filter of LMS to collecting the data waveform obtained；Using the data after treatment classification as training sample, generation is learnt 60 years old and the pervasive health model of the above crowd using LSTM network trainings；For existing pervasive health model, after obtaining personal data, model is identified while continuous training and personalization adjust, and makes the abnormity early warning of corresponding situation.The present invention organically combines conventional statistics model algorithm and neural network, and to 60 years old and the above crowd carried out health monitoring in all directions, realizes the personalized health prediction and early warning of efficiently and accurately, and lay a good foundation is established to establish the endowment service of wisdom health.
Description
Technical field
The invention belongs to the technical fields such as neural network, deep learning, Internet of Things, wisdom endowment, concretely relate to
A kind of doctor based on LSTM networks is foster to combine health monitor method.
Background technology
During wisdom is supported parents and developed, for endowment Service Source complexity dispersion, doctor's nutrient is from can not provide individual character
The problems such as change, active, service of safe, accurate, lasting Life cycle, research transboundary services the wisdom for curing and supporting and combining more and is good for
Health old-age provision model is just in gradual perfection.By integrating wearable, portable, contactless health monitoring and smart home device,
It establishes and the healthy in all directions of services such as looks after towards chronic diseases management, the first aid of old burst disease, Disease Warning Mechanism, critical old man
Internet of Things is monitored, monitors aged health physical signs in real time, the collection and utilization of total health information is realized, is carried with HIS systems
Old man's Electronic Health Record of confession, medical diagnosis and treat data etc. merge, and form health monitoring large data center, based on big number
Health forecast and early warning are realized according to analytical technology, and medical treatment auxiliary, health account inquiry, health Evaluation report, health are provided
The services such as assistant, health knowledge.
During improving the wisdom health old-age provision model for transboundary servicing more and curing and supporting and combine, realization health forecast and early warning
It is one of key technology.By to 60 years old and the collection of the total health information of the above crowd, including vital signs in real time number
According to information is carried out processing analysis mining, finds rule therein or become by medical history, the medication history etc. of 60 years old and the above crowd
Gesture can be both carried out with additional medical facility to 60 years old and early warning that the health status of the above crowd is more accurately predicted
Treatment, and can help 60 years old and the above crowd and relatives understand health condition, to 60 years old and the above crowd took specific aim to control
It treats and looks after.To evade privacy, avoid by the way of image monitoring, big data analysis technology becomes mainstream.For long period
The analysis of dynamic time sequence data has long-term regular feature, and needs to carry out a degree of prediction, common data
Excavation mode cannot be satisfied demand, such as clustering etc., therefore select a kind of deformation using RNN in neural network, LSTM
(Long Short-Term Memory) network, i.e. shot and long term memory network.LSTM networks are a kind of time recurrent neural networks,
It is suitable for being spaced and postpone relatively long critical event in processing and predicted time sequence, is suitable for required health forecast
With early warning demand.
Invention content
It is supported the purpose of the present invention is to provide a kind of doctor based on LSTM networks and combines health monitor method, kept away with realizing
In the case of exempting from invasion of privacy, to 60 years old and the above crowd carried out health monitoring in all directions, and health is realized by big data analysis
Prediction and early warning the services such as look after to establish towards chronic diseases management, the first aid of old burst disease, Disease Warning Mechanism, critical old man
Lay a good foundation is established, the target of health endowment is reached.
The present invention proposes that a kind of doctor based on LSTM networks supports and combines health monitor method, and this method includes walking in detail below
Suddenly：
Step 1：Health detection master data is collected, cloud database is stored in；
Step 2：Pretreatment and noise reduction are carried out to the data being collected into；
Step 3：To treated data carry out conclusion training, generate 60 years old and the pervasive health model of the above crowd；
Step 4：Personalized model study is carried out using LSTM networks to pervasive health model, generates 60 years old and the above crowd
Individual personalized health model persistently carries out model adjustment and identification, and makes the abnormity early warning of corresponding situation.
In the specific steps, the first step is characterized in, the collection health detection master data refers to：Via health
Detection device, smart home and HIS systems collect the health letter in all directions for obtaining at least 10,000 60 years old and the above crowds
Cease data, including full vital sign data, life data, diagnosis and treatment data, environmental data and other multidimensional behavioral datas, wherein
Full vital sign data includes heart rate, pulse, blood pressure, blood oxygen, body temperature, breathing, and life data include the life rule in sequential
Rule, real time position, diagnosis and treatment data include medical history, medication history, allergies, history of operation, and environmental data includes temperature (environment temperature
And sendible temperature), humidity, composition of air, intensity of illumination (ultraviolet index), other multidimensional behavioral datas include pressure sensing
The data that device and infrared sensor provide；
In the specific steps, second step is characterized in, carries out pretreatment and noise reduction to the data being collected into, specially：
The data waveform of health and fitness information in all directions to collecting 60 years old and the above crowd obtaining is carried out using the adaptive notch filter of LMS
Denoising：
Assuming that signal is superimposed for mono-tone interference, LMS adaptive notch filter denoisings are：
Wherein x (t) is input signal, and s (t) is signal after denoising,For noise signal；
Sampling, i.e., it is required using 60 years old and the data waveform of health and fitness information in all directions of the above crowd is as input signal d (n)
Denoising after signal be s (n), so that the weight coefficient of linear combination is adaptively adjusted by LMS algorithm, to make error
Signal is minimum and goes to zero, and realizes the purpose for eliminating interference denoising：
Wherein,
Wherein f0For sample frequency, fsFor former data acquiring frequency；
Afterwards by data according to full vital sign data, life data, diagnosis and treatment data, environmental data and other multidimensional behavior numbers
According to being divided into five kinds of data categories and preserving, as model training data sample, each of which data record includes serial number, institute
Category people number, time, data category, data represent the fields such as meaning, data content, data unit；
In the specific steps, third step is characterized in that data carry out conclusion training by treated, specially：It will place
It manages sorted 60 years old and the full vital sign data of the above crowd, life data, diagnosis and treatment data, environmental data and other are more
Tie up behavioral data as training data sample, by 60 years old and the above crowd it is daily activity there are many situation, including morning exercises,
Breakfast reads newspaper, sees TV, lunch, afternoon nap, stroll, dinner, sleep, and totally 9 class, does time-slotting feature extraction, i.e., by these
Situation, which labels, carries out the feature extraction based on statistical model and classification, the feature dimensions for solving to be likely to occur using collaborative filtering
Spend less or the more problem of duplicate data, then generated 60 years old using the study of LSTM network trainings and the above crowd it is pervasive strong
Health model, in LSTM network training process：
The design for carrying out network first, converts each sample of training data to 1024 dimension row vectors, input sample
First pass through one layer of LSTM network and one layer 50% of Dropout, then by one layer of LSTM network and one layer 50% of Dropout,
Data fitting is finally done by one layer of LSTM network again, Dropout layers are clipped between every two layers of LSTM networks to prevent over-fitting,
Then the fitting of multiple data is integrated, is classified by full articulamentum, be finally classified as one kind, is obtained 60 years old or more
The pervasive health model of crowd；
Learning rate variation can be obtained after training, wherein training set discrimination converges to 97%, and test machine discrimination is
90.5%.
In the specific steps, the 4th step is characterized in, the adjustment of individual character model and identification, specially：For existing general
Suitable health model after obtaining personal data, is identified model while continuous training adjusts, constantly by model according to individual
Actual conditions are adjusted to personalization.This include full vital sign data, life data, diagnosis and treatment data, environmental data and its
His multidimensional behavioral data carries out the study of personalized model using LSTM networks, to generate 60 years old and above crowd individual
Property health model, this model is continued to optimize in identification process, while improving recognition accuracy；It is obtained by training pattern
This is identified the predicted value of every health data of individual, obtains individual real time position to determine environmental form where individual, will
The predicted value of individual primary health data and the environmental index range of place environmental form are used as the standard of referring to, and will monitor in real time
The every health data of individual and environmental data be compared with term of reference, it is abnormal with environmental abnormality item to provide individual health
Reference value, and make the abnormity early warning of corresponding situation.
Implement a kind of doctor based on LSTM networks provided by the invention to support in conjunction with health monitor method compared with prior art
It has the advantages that：The present invention was cast aside to 60 years old and above crowd's individual carries out the limitation of image monitoring, by LSTM technologies
Applied to 60 years old and above crowd's long duration health monitoring and abnormity early warning to a certain extent in all directions, when increasing
Between influence of the priority factor to recognition result, be very suitable for the important data set of this time order and function factor of health monitoring early warning,
To improve discrimination；The characteristics of reference face recognition algorithms of novelty, carries out model learning using first extraction generic features
Then training carries out lasting for a long time model according to personalized real data situation and adjusts, with adapt to the feature of user to
Amount provides health monitoring and Warning Service more precisely, personalized for user.In addition, the present invention adopts for the first time on the implementation
Use paddle as training frame, which is domestic frame, is suitble to use in engineering.
Description of the drawings
Fig. 1 is the flow chart of the present invention.
Specific implementation mode
Technical solution in the embodiment of the present invention is done into further clear, perfect be described in detail below.Obviously, this hair
The embodiment that bright described embodiment is not all of, based on the embodiments of the present invention, persons skilled in the art are not having
There is the every other embodiment obtained under the premise of making creative work, shall fall within the protection scope of the present invention.
It is supported the present invention provides a kind of doctor based on LSTM networks and combines health monitor method, to health detection equipment, intelligence
Can collect 60 years old of household and HIS systems and the above crowd health and fitness information data carry out learning model building in all directions, using LSTM
(Long Short-Term Memory) network method, to realize in the case where avoiding invasion of privacy, to 60 years old and the above people
Group carries out the long-term monitoring of personalized health in all directions, analysis prediction and a degree of early warning.
The basic data source of the present invention is to be capable of providing full vital sign data, life data, diagnosis and treatment data, environment number
According to and other multidimensional behavioral datas health detection equipment, smart home and HIS systems, wherein full vital sign data packet
Include heart rate, pulse, blood pressure, blood oxygen, body temperature, breathing, life data include rule of life in sequential, real time position, diagnosis and treatment number
According to including medical history, medication history, allergies, history of operation, environmental data includes temperature (environment temperature and sendible temperature), humidity, sky
Gas ingredient, intensity of illumination (ultraviolet index), other multidimensional behavioral datas include that pressure sensor and infrared sensor provide
Data
To keep objects, features and advantages of the present invention more clear and easy to understand, With reference to embodiment to this
Invention is described in further detail.
The present invention proposes that a kind of doctor based on LSTM networks supports and combines health monitor method, and this method includes walking in detail below
Suddenly：
Step 1：Health detection master data is collected, cloud database is stored in；
Step 2：Pretreatment and noise reduction are carried out to the data being collected into；
Step 3：To treated data carry out conclusion training, generate 60 years old and the pervasive health model of the above crowd；
Step 4：Personalized model study is carried out using LSTM networks to pervasive health model, generates 60 years old and the above crowd
Individual personalized health model persistently carries out model adjustment and identification, and makes the abnormity early warning of corresponding situation.
In the specific steps, the first step is characterized in, the collection health detection master data refers to：Via health
Detection device, smart home and HIS systems collect the health letter in all directions for obtaining at least 10,000 60 years old and the above crowds
Cease data, including full vital sign data, life data, diagnosis and treatment data, environmental data and other multidimensional behavioral datas, wherein
Full vital sign data includes heart rate, pulse, blood pressure, blood oxygen, body temperature, breathing, and life data include the life rule in sequential
Rule, real time position, diagnosis and treatment data include medical history, medication history, allergies, history of operation, and environmental data includes temperature (environment temperature
And sendible temperature), humidity, composition of air, intensity of illumination (ultraviolet index), other multidimensional behavioral datas include pressure sensing
The data that device and infrared sensor provide；Health monitoring master data is stored in Ali's cloud, in case follow-up training process uses.
In the specific steps, second step is characterized in, carries out pretreatment and noise reduction to the data being collected into, specially：
The data waveform of health and fitness information in all directions to collecting 60 years old and the above crowd obtaining is carried out using the adaptive notch filter of LMS
Denoising：
Assuming that signal is superimposed for mono-tone interference, LMS adaptive notch filter denoisings are：
Wherein x (t) is input signal, and s (t) is signal after denoising,For noise signal；
Sampling, i.e., it is required using 60 years old and the data waveform of health and fitness information in all directions of the above crowd is as input signal d (n)
Denoising after signal be s (n), so that the weight coefficient of linear combination is adaptively adjusted by LMS algorithm, to make error
Signal is minimum and goes to zero, and realizes the purpose for eliminating interference denoising：
Wherein,
Wherein f0For sample frequency, fsFor former data acquiring frequency；
Afterwards by data according to full vital sign data, life data, diagnosis and treatment data, environmental data and other multidimensional behavior numbers
According to being divided into five kinds of data categories and preserving, as model training data sample, each of which data record includes serial number, institute
Category people number, time, data category, data represent the fields such as meaning, data content, data unit；
Specific data record format is as follows：
1 full vital sign data format of table
The life data format of table 2
3 diagnosis and treatment data format of table
4 environmental data format of table
Other multidimensional behavioral data formats of table 5
In the specific steps, third step is characterized in that data carry out conclusion training by treated, specially：It will place
It manages sorted 60 years old and the full vital sign data of the above crowd, life data, diagnosis and treatment data, environmental data and other are more
Tie up behavioral data as training data sample, by 60 years old and the above crowd it is daily activity there are many situation, including morning exercises,
Breakfast reads newspaper, sees TV, lunch, afternoon nap, stroll, dinner, sleep, and totally 9 class, does time-slotting feature extraction, i.e., by these
Situation, which labels, carries out the feature extraction based on statistical model and classification, the feature dimensions for solving to be likely to occur using collaborative filtering
Spend less or the more problem of duplicate data, then generated 60 years old using the study of LSTM network trainings and the above crowd it is pervasive strong
Health model.
Wherein, in annotation process, the data of all the sensors are arranged according to timestamps ordering, for different 60 years old
And above crowd's individual, different time sections, different movement tracks, different animations, by data according to period contingency table
Note.These action are classified as morning exercises, morning by the data passed back according to all the sensors, the action of the tested old man of accurate reproduction
It eats, read newspaper, seeing TV, lunch, afternoon nap, stroll, dinner, sleep, the data block of totally 9 different labels, makees not different labels
New label field is added to be stored into database simultaneously in data with mark.
In LSTM network training process, the training environment of use and it is expected that duration is as follows：
Equipment and environment：Server 1, double Teslak80 video cards, 7 systems of CentOS, 500G (SSD) Ali's cloud
Training frame：Paddlepaddle
Use language：Python
It is expected that training duration：8 days
After obtaining data, data is normalized (i.e. by all aggregation of data to 0 to 1 range, passes through first
All data divided by data maximums are obtained), the design of deep learning frame is then carried out, by each of training data
Sample is converted into 1024 dimension row vectors, and input sample first passes through one layer of LSTM network and one layer 50% of Dropout, then passes through
One layer of LSTM network and one layer 50% of Dropout finally do data fitting, Dropout layers of folder by one layer of LSTM network again
To prevent over-fitting between every two layers of LSTM networks, then the fitting of multiple data is integrated, is divided by full articulamentum
Class is finally classified as one kind, can obtain 60 years old and the pervasive health model of the above crowd.
Training sample is imported in the deep learning frame built with Paddlepaddle, is voluntarily trained by machine, repeatedly
Generation 200 times, relatively convergent training result is finally obtained, wherein training set discrimination converges to 97%, and test machine discrimination is
90.5%.
In the specific steps, the 4th step is characterized in, the adjustment of individual character model and identification, specially：For existing general
Suitable health model after obtaining personal data, is identified model while continuous training adjusts, constantly by model according to individual
Actual conditions are adjusted to personalization.This include full vital sign data, life data, diagnosis and treatment data, environmental data and its
His multidimensional behavioral data carries out the study of personalized model using LSTM networks, to generate 60 years old and above crowd individual
Property health model, this model is continued to optimize in identification process, while improving recognition accuracy；It is obtained by training pattern
This is identified the predicted value of every health data of individual, obtains individual real time position to determine environmental form where individual, will
The predicted value of individual primary health data and the environmental index range of place environmental form are used as the standard of referring to, and will monitor in real time
The every health data of individual and environmental data be compared with term of reference, it is abnormal with environmental abnormality item to provide individual health
Reference value, and make the abnormity early warning of corresponding situation.
The process configuration environment and to estimate duration as follows：
Use model：Obtained training pattern, volume are about 150MB
Equipment and environment：Server 1, double Teslak80 video cards, 7 systems of CentOS, 500G (SSD) Ali's cloud
Database：MySQL 5.7
Web frames：Tomcat 8
The Refresh Data time：300s
Duration is estimated in calculating：10ms.
Process is：For existing pervasive health model, personal data are constantly obtained, this includes full vital sign number
According to, life data, diagnosis and treatment data, environmental data and other multidimensional behavioral datas, data are normalized, data
Once refreshed per 300s, the data of 3600s before acquisition are exported to which result be calculated by monitor.
Such as there are the data (this decision threshold is 50%) that result differs markedly from sample pattern, then alarms, by 60
The specific residence and room output of year and above crowd individual are accurately positioned.Specific implementation is：
A) from sensors for data, it is inputted into database, is arranged in timestamp order；
B) noise reduction is integrally carried out to data, will drift about and larger data screening and revise in data, to allow data as possible
Neatly；
C) by reduced data, input model obtains result via model training；
D) test result is returned, obtains the corresponding life pattern of prediction and prediction probability；
E) prediction threshold value is set and time threshold, threshold value are compared with prediction probability, often persistently 600s predicted values are less than
Threshold value then assert that appearance is abnormal, and severe patient does emergency response processing；
F) the database data of update in every 3 minutes, comes back to a) after update.
Method and step described in conjunction with the examples disclosed in this document, which can configure, builds programmed environment to implement.It can be with
Understand, for those of ordinary skill in the art, can be conceived with the technique according to the invention and make various other phases
The change and deformation answered, and all these changes and deformation should all belong to the protection domain of the claims in the present invention.To sum up institute
It states, the content of the present specification should not be construed as limiting the invention.
Claims (5)
1. a kind of doctor based on LSTM networks, which supports, combines health monitor method, which is characterized in that this approach includes the following steps：
Step 1：Health detection master data is collected, cloud database is stored in；
Step 2：Pretreatment and noise reduction are carried out to the data being collected into；
Step 3：To treated data carry out conclusion training, generate 60 years old and the pervasive health model of the above crowd；
Step 4：Personalized model study is carried out using LSTM networks to pervasive health model, generates 60 years old and the above crowd is individual
Personalized health model persistently carries out model adjustment and identification, and makes the abnormity early warning of corresponding situation.
2. doctor according to claim 1, which supports, combines health monitor method, which is characterized in that the step 1 specifically includes：Through
By health detection equipment, smart home and HIS systems, collects and obtain being good in all directions at least 10,000 60 years old and the above crowds
Health information data, including full vital sign data, life data, diagnosis and treatment data, environmental data and other multidimensional behavioral datas,
In, full vital sign data includes heart rate, pulse, blood pressure, blood oxygen, body temperature, breathing；Life data include the life rule in sequential
Rule, real time position；Diagnosis and treatment data include medical history, medication history, allergies, history of operation；Environmental data includes temperature, humidity, air
Ingredient, intensity of illumination, other multidimensional behavioral datas include the data that pressure sensor and infrared sensor provide.
3. doctor according to claim 1, which supports, combines health monitor method, which is characterized in that the step 2 specifically includes：It is right
The data waveform of health and fitness information in all directions for collecting 60 years old and the above crowd obtaining carries out denoising using the adaptive notch filter of LMS
Processing：
Assuming that signal is superimposed for mono-tone interference, LMS adaptive notch filter denoisings are：
Wherein x (t) is input signal, and s (t) is signal after denoising,For noise signal；
Sampling, i.e., using 60 years old and the data waveform of health and fitness information in all directions of the above crowd is as input signal d (n), required goes
Signal is s (n) after making an uproar, and so that the weight coefficient of linear combination is adaptively adjusted by LMS algorithm, to make error signal pole
It is small and go to zero, realize the purpose for eliminating interference denoising：
Wherein,
Wherein f0For sample frequency, fsFor former data acquiring frequency；
Afterwards by data according to full vital sign data, life data, diagnosis and treatment data, environmental data and other multidimensional behavioral datas point
It at five kinds of data categories and preserves, as model training data sample, each of which data record includes serial number, affiliated people volume
Number, time, data category, data represent meaning, data content, data unit field.
4. doctor according to claim 1, which supports, combines health monitor method, which is characterized in that the step 3 specifically includes：It will
The full vital sign data of 60 years old after treatment classification and the above crowd, life data, diagnosis and treatment data, environmental data and other are more
Behavioral data is tieed up as training data sample, situation, including morning exercises, morning by 60 years old and there are many daily activities of the above crowd
It eats, read newspaper, seeing TV, lunch, afternoon nap, stroll, dinner, sleep, totally 9 class, doing time-slotting feature extraction, i.e., by these situations
It labels and carries out the feature extraction based on statistical model and classification, it is very few to solve the characteristic dimension being likely to occur using collaborative filtering
Or the problem that duplicate data is more, then generation is learnt 60 years old and the pervasive health model of the above crowd using LSTM network trainings；
Wherein, in LSTM network training process：
Network design is carried out first, converts each sample of training data to 1024 dimension row vectors, input sample first passes through
One layer of LSTM network and one layer 50% of Dropout, then by one layer of LSTM network and one layer 50% of Dropout, finally again
Data fitting is done by one layer of LSTM network, Dropout layers are clipped between every two layers of LSTM networks to prevent over-fitting, then will be more
The fitting of a data is integrated, and is classified by full articulamentum, is finally classified as one kind, obtain 60 years old and the above crowd it is pervasive
Health model；Learning rate variation is obtained after training, wherein training set discrimination converges to 97%, and test machine discrimination is
90.5%.
5. doctor according to claim 1, which supports, combines health monitor method, which is characterized in that the step 4 specifically includes：It is right
In existing pervasive health model, after obtaining personal data, model is identified while continuous training adjusts, constantly by model
It is adjusted to personalization according to personal actual conditions；This includes full vital sign data, life data, diagnosis and treatment data, environment number
According to this and other multidimensional behavioral datas, the study that personalized model is carried out using LSTM networks, to generate 60 years old and the above people
The individual personalized health model of group, this model is continued to optimize in identification process, while improving recognition accuracy；Pass through training mould
Type show that this is identified the predicted value of every health data of individual, obtains individual real time position to determine environmental classes where individual
Type will be real-time using the environmental index range of the predicted value of individual primary health data and place environmental form as the standard of referring to
The every health data of individual and environmental data of monitoring are compared with term of reference, provide individual health exception and environmental abnormality
The reference value of item, and make the abnormity early warning of corresponding situation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810417954.0A CN108597609A (en) | 2018-05-04 | 2018-05-04 | A kind of doctor based on LSTM networks is foster to combine health monitor method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810417954.0A CN108597609A (en) | 2018-05-04 | 2018-05-04 | A kind of doctor based on LSTM networks is foster to combine health monitor method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108597609A true CN108597609A (en) | 2018-09-28 |
Family
ID=63619811
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810417954.0A Pending CN108597609A (en) | 2018-05-04 | 2018-05-04 | A kind of doctor based on LSTM networks is foster to combine health monitor method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108597609A (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109447164A (en) * | 2018-11-01 | 2019-03-08 | 厦门大学 | A kind of motor behavior method for classifying modes, system and device |
| CN109472311A (en) * | 2018-11-13 | 2019-03-15 | 北京物灵智能科技有限公司 | A kind of user behavior recognition method and device |
| CN109615012A (en) * | 2018-12-13 | 2019-04-12 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109636421A (en) * | 2018-12-13 | 2019-04-16 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109659035A (en) * | 2018-12-13 | 2019-04-19 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109685671A (en) * | 2018-12-13 | 2019-04-26 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109767836A (en) * | 2018-12-29 | 2019-05-17 | 上海亲看慧智能科技有限公司 | A kind of medical diagnosis artificial intelligence system, device and its self-teaching method |
| CN110458355A (en) * | 2019-08-09 | 2019-11-15 | 泰康保险集团股份有限公司 | Event prediction method, apparatus, equipment and storage medium |
| CN110827994A (en) * | 2020-01-13 | 2020-02-21 | 四川大学华西医院 | Myocardial infarction early warning method, device and equipment and storage medium |
| CN111026004A (en) * | 2019-11-07 | 2020-04-17 | 来邦科技股份公司 | Environment monitoring method and system |
| CN111428864A (en) * | 2020-04-08 | 2020-07-17 | 重庆软汇科技股份有限公司 | Automatic transaction generation method and system based on neural network |
| CN111616695A (en) * | 2020-06-29 | 2020-09-04 | 歌尔科技有限公司 | Heart rate acquisition method, device, system and medium |
| WO2020207317A1 (en) * | 2019-04-09 | 2020-10-15 | Oppo广东移动通信有限公司 | User health assessment method and apparatus, and storage medium and electronic device |
| CN111798980A (en) * | 2020-07-10 | 2020-10-20 | 哈尔滨工业大学（深圳） | Complex medical biological signal processing method and device based on deep learning network |
| CN111951965A (en) * | 2020-07-31 | 2020-11-17 | 合肥工业大学 | Panoramic health dynamic monitoring and predicting system based on time sequence knowledge graph |
| CN112037932A (en) * | 2020-09-09 | 2020-12-04 | 平安科技（深圳）有限公司 | Patient medication behavior intervention method and device, server and storage medium |
| CN112257608A (en) * | 2020-10-23 | 2021-01-22 | 成都希盟泰克科技发展有限公司 | Yak breeding health state monitoring method |
| CN112259260A (en) * | 2020-11-18 | 2021-01-22 | 中国科学院自动化研究所 | Intelligent medical question and answer method, system and device based on intelligent wearable equipment |
| CN112289430A (en) * | 2020-09-30 | 2021-01-29 | 音数汇元(上海)智能科技有限公司 | Home care information pushing method and device, computer equipment and storage medium |
| CN112735609A (en) * | 2020-12-31 | 2021-04-30 | 浙江禾连网络科技有限公司 | LSTM model-based old people biological signal health monitoring method |
| CN113220801A (en) * | 2021-05-17 | 2021-08-06 | 支付宝(杭州)信息技术有限公司 | Structured data classification method, device, equipment and medium |
| CN114999649A (en) * | 2022-06-06 | 2022-09-02 | 广州培生智能科技有限公司 | Old people physical sign data monitoring and early warning method and system based on deep learning |
| WO2023226369A1 (en) * | 2022-05-26 | 2023-11-30 | 深圳先进技术研究院 | Method, apparatus and device for predicting weaning from mechanical ventilation, and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106991506A (en) * | 2017-05-16 | 2017-07-28 | 深圳先进技术研究院 | Intelligent terminal and its stock trend forecasting method based on LSTM |
| CN106999111A (en) * | 2014-10-01 | 2017-08-01 | 纽洛斯公司 | System and method for detecting invisible human emotion |
| CN107491638A (en) * | 2017-07-28 | 2017-12-19 | 深圳和而泰智能控制股份有限公司 | A kind of ICU user's prognosis method and terminal device based on deep learning model |
| CN107506602A (en) * | 2017-09-07 | 2017-12-22 | 北京海融兴通信息安全技术有限公司 | A kind of big data health forecast system |
| CN107769972A (en) * | 2017-10-25 | 2018-03-06 | 武汉大学 | A kind of power telecom network equipment fault Forecasting Methodology based on improved LSTM |
| CN107786369A (en) * | 2017-09-26 | 2018-03-09 | 广东电网有限责任公司电力调度控制中心 | Based on the perception of IRT step analyses and LSTM powerline network security postures and Forecasting Methodology |
| CN107864405A (en) * | 2017-11-14 | 2018-03-30 | 国家新闻出版广电总局广播科学研究院 | A kind of Forecasting Methodology, device and the computer-readable medium of viewing behavior type |
-
2018
- 2018-05-04 CN CN201810417954.0A patent/CN108597609A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106999111A (en) * | 2014-10-01 | 2017-08-01 | 纽洛斯公司 | System and method for detecting invisible human emotion |
| CN106991506A (en) * | 2017-05-16 | 2017-07-28 | 深圳先进技术研究院 | Intelligent terminal and its stock trend forecasting method based on LSTM |
| CN107491638A (en) * | 2017-07-28 | 2017-12-19 | 深圳和而泰智能控制股份有限公司 | A kind of ICU user's prognosis method and terminal device based on deep learning model |
| CN107506602A (en) * | 2017-09-07 | 2017-12-22 | 北京海融兴通信息安全技术有限公司 | A kind of big data health forecast system |
| CN107786369A (en) * | 2017-09-26 | 2018-03-09 | 广东电网有限责任公司电力调度控制中心 | Based on the perception of IRT step analyses and LSTM powerline network security postures and Forecasting Methodology |
| CN107769972A (en) * | 2017-10-25 | 2018-03-06 | 武汉大学 | A kind of power telecom network equipment fault Forecasting Methodology based on improved LSTM |
| CN107864405A (en) * | 2017-11-14 | 2018-03-30 | 国家新闻出版广电总局广播科学研究院 | A kind of Forecasting Methodology, device and the computer-readable medium of viewing behavior type |
Non-Patent Citations (1)
| Title |
|---|
| 陈智: "一种基于LMS算法自适应陷波器的设计", 《郧阳师范高等专科学校学报》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109447164A (en) * | 2018-11-01 | 2019-03-08 | 厦门大学 | A kind of motor behavior method for classifying modes, system and device |
| CN109472311A (en) * | 2018-11-13 | 2019-03-15 | 北京物灵智能科技有限公司 | A kind of user behavior recognition method and device |
| CN109615012A (en) * | 2018-12-13 | 2019-04-12 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109636421A (en) * | 2018-12-13 | 2019-04-16 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109659035A (en) * | 2018-12-13 | 2019-04-19 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109685671A (en) * | 2018-12-13 | 2019-04-26 | 平安医疗健康管理股份有限公司 | Medical data exception recognition methods, equipment and storage medium based on machine learning |
| CN109767836A (en) * | 2018-12-29 | 2019-05-17 | 上海亲看慧智能科技有限公司 | A kind of medical diagnosis artificial intelligence system, device and its self-teaching method |
| WO2020207317A1 (en) * | 2019-04-09 | 2020-10-15 | Oppo广东移动通信有限公司 | User health assessment method and apparatus, and storage medium and electronic device |
| CN110458355A (en) * | 2019-08-09 | 2019-11-15 | 泰康保险集团股份有限公司 | Event prediction method, apparatus, equipment and storage medium |
| CN111026004A (en) * | 2019-11-07 | 2020-04-17 | 来邦科技股份公司 | Environment monitoring method and system |
| CN110827994A (en) * | 2020-01-13 | 2020-02-21 | 四川大学华西医院 | Myocardial infarction early warning method, device and equipment and storage medium |
| CN110827994B (en) * | 2020-01-13 | 2020-04-17 | 四川大学华西医院 | Myocardial infarction early warning method, device and equipment and storage medium |
| CN111428864A (en) * | 2020-04-08 | 2020-07-17 | 重庆软汇科技股份有限公司 | Automatic transaction generation method and system based on neural network |
| CN111616695A (en) * | 2020-06-29 | 2020-09-04 | 歌尔科技有限公司 | Heart rate acquisition method, device, system and medium |
| CN111798980A (en) * | 2020-07-10 | 2020-10-20 | 哈尔滨工业大学（深圳） | Complex medical biological signal processing method and device based on deep learning network |
| CN111951965A (en) * | 2020-07-31 | 2020-11-17 | 合肥工业大学 | Panoramic health dynamic monitoring and predicting system based on time sequence knowledge graph |
| CN111951965B (en) * | 2020-07-31 | 2024-01-23 | 合肥工业大学 | Panoramic health dynamic monitoring and predicting system based on time sequence knowledge graph |
| WO2021151329A1 (en) * | 2020-09-09 | 2021-08-05 | 平安科技（深圳）有限公司 | Patient medication behavior intervention method and apparatus, and server and storage medium |
| CN112037932A (en) * | 2020-09-09 | 2020-12-04 | 平安科技（深圳）有限公司 | Patient medication behavior intervention method and device, server and storage medium |
| CN112289430A (en) * | 2020-09-30 | 2021-01-29 | 音数汇元(上海)智能科技有限公司 | Home care information pushing method and device, computer equipment and storage medium |
| CN112257608A (en) * | 2020-10-23 | 2021-01-22 | 成都希盟泰克科技发展有限公司 | Yak breeding health state monitoring method |
| CN112259260A (en) * | 2020-11-18 | 2021-01-22 | 中国科学院自动化研究所 | Intelligent medical question and answer method, system and device based on intelligent wearable equipment |
| CN112259260B (en) * | 2020-11-18 | 2023-11-17 | 中国科学院自动化研究所 | Intelligent medical question-answering method, system and device based on intelligent wearable equipment |
| CN112735609A (en) * | 2020-12-31 | 2021-04-30 | 浙江禾连网络科技有限公司 | LSTM model-based old people biological signal health monitoring method |
| CN113220801A (en) * | 2021-05-17 | 2021-08-06 | 支付宝(杭州)信息技术有限公司 | Structured data classification method, device, equipment and medium |
| CN113220801B (en) * | 2021-05-17 | 2022-07-29 | 支付宝(杭州)信息技术有限公司 | Structured data classification method, device, equipment and medium |
| WO2023226369A1 (en) * | 2022-05-26 | 2023-11-30 | 深圳先进技术研究院 | Method, apparatus and device for predicting weaning from mechanical ventilation, and storage medium |
| CN114999649A (en) * | 2022-06-06 | 2022-09-02 | 广州培生智能科技有限公司 | Old people physical sign data monitoring and early warning method and system based on deep learning |
| CN114999649B (en) * | 2022-06-06 | 2022-12-27 | 广州培生智能科技有限公司 | Old people sign data monitoring and early warning method and system based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108597609A (en) | A kind of doctor based on LSTM networks is foster to combine health monitor method | |
| Umematsu et al. | Improving students' daily life stress forecasting using LSTM neural networks | |
| Wang et al. | Human activity recognition in a smart home environment with stacked denoising autoencoders | |
| Zhu et al. | A deep learning approach for recognizing activity of daily living (ADL) for senior care: Exploiting interaction dependency and temporal patterns | |
| CN109949936A (en) | A kind of Risk Forecast Method of being hospitalized again based on deep learning mixed model | |
| Yang et al. | Coronary heart disease optimization system on adaptive-network-based fuzzy inference system and linear discriminant analysis (ANFIS–LDA) | |
| Chen et al. | Latent feature learning for activity recognition using simple sensors in smart homes | |
| Morid et al. | Learning hidden patterns from patient multivariate time series data using convolutional neural networks: A case study of healthcare cost prediction | |
| El-Attar et al. | Discrete wavelet transform-based freezing of gait detection in Parkinson’s disease | |
| Urwyler et al. | Recognition of activities of daily living in healthy subjects using two ad-hoc classifiers | |
| CN111248859A (en) | Automatic sleep apnea detection method based on convolutional neural network | |
| Lawanont et al. | Daily stress recognition system using activity tracker and smartphone based on physical activity and heart rate data | |
| Rahman et al. | Improving alcoholism diagnosis: comparing instance-based classifiers against neural networks for classifying EEG signal | |
| Thilakarathne et al. | Artificial Intelligence-Enabled IoT for Health and Wellbeing Monitoring | |
| CN114191665A (en) | Method and device for classifying man-machine asynchronous phenomena in mechanical ventilation process | |
| CN109119155B (en) | ICU death risk assessment system based on deep learning | |
| Xu et al. | Decode brain system: A dynamic adaptive convolutional quorum voting approach for variable-length EEG data | |
| Malviya et al. | Mental stress level detection using LSTM for WESAD dataset | |
| Jain et al. | Prediction of Stress using Machine Learning and IoT | |
| Xu et al. | Comparative studies on activity recognition of elderly people living alone | |
| Gao et al. | Human–System Interaction Based on Eye Tracking for a Virtual Workshop | |
| Dhanusha et al. | Enhanced deep hierarchical classification model for smart home-based alzheimer disease detection | |
| Lu et al. | Self-learning based motion recognition using sensors embedded in a smartphone for mobile healthcare | |
| Kostek et al. | Knowledge representation of motor activity of patients with Parkinson’s disease | |
| Hao et al. | Feature Optimization of Exhaled Breath Signals Based on Pearson‐BPSO |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180928 |
|
| RJ01 | Rejection of invention patent application after publication |