CN104346369A - Method of establishing heartbeat impulse wave form feature library - Google Patents
Method of establishing heartbeat impulse wave form feature library Download PDFInfo
- Publication number
- CN104346369A CN104346369A CN201310327012.0A CN201310327012A CN104346369A CN 104346369 A CN104346369 A CN 104346369A CN 201310327012 A CN201310327012 A CN 201310327012A CN 104346369 A CN104346369 A CN 104346369A
- Authority
- CN
- China
- Prior art keywords
- heartbeat
- vector
- wave form
- shock wave
- feature database
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
- G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Signal Processing (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
The invention provides a method of establishing a heartbeat impulse wave form feature library. The method includes: selecting and adding a standard sample vector; adding new feature vectors by means of machine learning; correcting fringe data. Preferably, the step of selecting and adding the standard sample vector includes: collecting human heartbeat impulse wave information by measuring vibrations of a bed; selecting a sample waveform from the collected heartbeat impulse wave information; selecting feature points for the sample waveform; converting the sample waveform into a sample vector; adding the sample vector to the heartbeat impulse wave form feature library. The method as mentioned above has the advantages that establishing the heartbeat impulse wave form feature library by means of machine learning is efficient and well accurate, and the use of optimization leads to simple waveform sample vectors and feature vectors and low calculated quantity.
Description
Technical field
The present invention relates to a kind of method setting up heartbeat shock wave form feature database, particularly relate to a kind of method utilizing the mode of machine learning to set up heartbeat shock wave form feature database.
Background technology
Medical science, healthy etc. numerous areas extensively needs to measure human heartbeat's feature, a kind of mode of the human heartbeat's of measurement feature be people lie on a bed static after, with shock sensor gather human heart beat time to bed produce vibration data thus obtain human heartbeat's feature (calling heartbeat shock wave morphological feature in the following text), although this mode facilitates and affects testee little, the data gathered can accuracy be poor because being subject to the impact of many factors.The factor of impact includes but not limited to: have people to pass by etc. around the posture that people lies on a bed, bed.So the heartbeat shock wave characteristic pattern data gathered in this way needs through screening or revises.At present shake mode that the heartbeat shock wave characteristic pattern data that obtains screens or revise for first to set up heartbeat shock wave form feature database to this by measuring bed, then the data measured in the heartbeat shock wave characteristic pattern data that obtains and storehouse are compared, if the gap measuring data in the data and storehouse obtained is greater than pre-determined threshold, then give up this data.The core of this method to set up high-quality heartbeat shock wave form feature database.But the method efficiency setting up this kind of heartbeat shock wave form feature database is at present low, poor accuracy, and calculated amount is large.
Summary of the invention
The shortcoming of prior art in view of the above, the object of the present invention is to provide a kind of method setting up heartbeat shock wave form feature database, for solve prior art set up jump that shock wave form feature database efficiency is low, poor accuracy and the large problem of calculated amount.
For achieving the above object and other relevant objects, the invention provides a kind of method setting up heartbeat shock wave form feature database, comprising: select and add master sample vector; New feature vector is added by the mode of machine learning; Revise marginal date.
Preferably, described selection the step of adding master sample vector comprise: gather people's heartbeat shockwave information by measuring bed vibrations; Sample waveform is selected from the described heartbeat shockwave information collected; For described sample waveform selects unique point; Described sample waveform is converted into sample vector; Described sample vector is added to described heartbeat shock wave form feature database.
As mentioned above, the method setting up heartbeat shock wave form feature database of the present invention, has following beneficial effect: set up heartbeat shock wave form feature database by the mode of machine learning, efficiency is high, and accuracy is good.And according to preferred mode, simply, calculated amount is little for the sample vector of waveform and proper vector.
Accompanying drawing explanation
Fig. 1 is shown as the process flow diagram setting up heartbeat shock wave form feature database method of the present invention;
Fig. 2 is shown as the static flat bed shocking waveshape figure recorded when crouching on bed of the measured;
Fig. 3 is shown as according to the design sketch of maps feature vectors to two-dimensional space setting up an embodiment generation of heartbeat shock wave form feature database method according to the present invention.
Embodiment
Below by way of specific instantiation, embodiments of the present invention are described, those skilled in the art the content disclosed by this instructions can understand other advantages of the present invention and effect easily.The present invention can also be implemented or be applied by embodiments different in addition, and the every details in this instructions also can based on different viewpoints and application, carries out various modification or change not deviating under spirit of the present invention.
Refer to accompanying drawing.It should be noted that, the diagram provided in the present embodiment only illustrates basic conception of the present invention in a schematic way, then only the assembly relevant with the present invention is shown in graphic but not component count, shape and size when implementing according to reality is drawn, it is actual when implementing, and the kenel of each assembly, quantity and ratio can be a kind of change arbitrarily, and its assembly layout kenel also may be more complicated.
Consult Fig. 1 below, Figure 1 shows that the present invention sets up the process flow diagram of the method for heartbeat shock wave form feature database.The method that present invention employs machine learning, to set up heartbeat shock wave form feature database, roughly can be divided into three steps, represent with step S1 ~ S3, describe in detail below to each step.
Step S1 represents and selects and add master sample.Selecting or add master sample can be manually complete, and also can be the data directly using other approach to obtain.In this example, adopt artificial selection and the mode of adding, concrete mode is: allow the lying on a bed of people's calmness to be measured, the heartbeat shock wave form signals that record sensor collection arrives, analyze the waveform obtaining heartbeat each time, this waveform is changed into sample vector, selects one or more groups sample vector of standard the most as master sample.This completes the input of the heartbeat shock wave form master sample that lies low.
Consult Fig. 2-3 below and the concrete grammar being obtained sample vector in this example by heartbeat surge waveform is described, for convenience of explanation, only lift flat crouching as example here.Fig. 2 shows and behaves static flat when crouching on bed, and the bed vibration data that shock sensor records, comprises heartbeat surge waveform in these data.First the unique point in waveform is selected.Maximum point J in waveform and minimal value K is chosen as unique point in this example.Then convert unique point to sample vector, during conversion, sample vector can be constructed by the amount of the position before characteristic feature point or mathematical relation.Only for convenience of explanation, the transformation rule of employing is this example: get the distance between a=J and K, and the slope of b=J and K 2 lines, then bivector (a, b) is sample vector.During actual use, in order to make result have more confidence level, desirable hyperspace, such as: the horizontal ordinate quadratic sum of c=J and K horizontal ordinate variance, d=J and K ..., then can obtain multidimensional sample vector be designated as (a, b, c, d ...).
In like manner, according to said method, can complete and sit up, the input of the sample that left/right is lain on one's side, for the purpose of simplifying the description, in this example, only input flat to be crouched, sit up, to be lain on the left side and crouch the sample input of four kinds of actions in right side, and a sample input is only set up in each action.Those skilled in the art also as required according to same other action of methods analyst, such as can go to bed, leave the bed, and also can set up the input of more sample to each action.For increasing data accuracy and more convenient to data separate from now on, in this example, by the identity information of measured personnel and action classification information (namely this shocking waveshape be measure people flat crouch, sit up, lie on the left side sleeping with right side or in going to bed or leaving the bed which action obtain) also together be added to heartbeat shock wave form feature database.Like this, if when needing from now on to utilize this heartbeat shock wave form feature database, such as, when needing the heartbeat data screening this certain action of personnel, just by the heartbeat shock wave form character separation of this action classification of these personnel when setting up this feature database out, screening accuracy can be increased further.In Fig. 3, the point of shade represents flat respectively from left to right, from top to bottom and crouches, sit up, lie on the left side and the mapping of sample vector in two-dimensional space crouched under four kinds of action classifications in right side.
It should be noted that, according to actual needs, in said method, those skilled in the art can adopt alternate manner to select unique point, or adopt alternate manner that waveform is converted into sample vector.Such as but not limited to: corresponding diagram 2 waveform, only choose more or less unique point, other polynomial function of corresponding employing generates sample vector; Or adopt the mode such as Fourier transform, wavelet transformation that waveform is changed into proper vector.
Because different people lies on a bed, the heartbeat shock wave form produced can not be identical, to this, can build a Sample Storehouse for everyone, then concluded by everyone Sample Storehouse in whole large heartbeat shock wave form feature database again.Thus the heartbeat shock wave form of different people can be recognized more accurately.
Step S2 represents and adds new feature vector by the mode of machine learning.Machine learning (Machine Learning) is exactly utilize computing machine to adopt conclusion, comprehensive mode, simulates or realizes the learning behavior of the mankind, reorganize the performance that the existing structure of knowledge makes it constantly to improve self.During machine learning, for the waveform newly collected, utilize calculating that itself and sample waveform are also changed into proper vector.After obtaining the proper vector of this waveform, by the two-dimensional space shown in this DUAL PROBLEMS OF VECTOR MAPPING to Fig. 3, then check that the vectorial sample vector from which block region of new feature is nearest, then think the heart shock wave form signals that the action representated by this region produces.And then calculate the similarity of this proper vector and this area sample vector, if similarity is higher than certain threshold value, then this proper vector is added heartbeat shock wave form feature database.
In this example, be simplified illustration, only to flat crouching, sitting up, lie on the left side and the sleeping four kinds of actions in right side respectively establish a sample input.For the heartbeat shock wave waveform newly collected, using point corresponding with J, K on sample waveform on it as unique point, and according to same mode in step S1, the heartbeat shock wave waveform newly collected is converted into proper vector.When judging characteristic vector is with the similarity of sample vector, this example adopts the mode of vector distance, calculate the distance of proper vector and sample orientation (if sample vector is multiple, then can calculate mean distance, proper vector and be positioned at the sample vector at center or be positioned at the distance etc. of sample vector at edge), if vector distance is less than certain threshold value, represent that similarity is high, then think and this new feature vector added heartbeat shock wave form feature database.The like, just can pass through machine learning mode, not stop to grow the sample inside feature database.Preferably, in this example, when adding new feature vector by the mode of machine learning, also measured's identity information and action classification information are added in heartbeat shock wave form feature database.
More preferably, in step s 2, machine can be set and differently learn in the different time periods, such as, at 0 ~ t
1time period, only calculate the distance of that sample vector in new feature vector and its region, if distance is less than threshold T
1, then this new feature vector is added heartbeat shock wave form feature database; And at t
1~ t
2time period, then by 0 ~ t
1the proper vector newly added in time period is also considered as sample vector, t
1~ t
2the interior new proper vector gathered needed the mean distance of sample vectors all in calculating and its region, if this mean distance is less than thresholding T time period
2, then new feature vector is added heartbeat shock wave form feature database, by that analogy.According to the present embodiment, after machine learning a period of time, the proper vector in heartbeat shock wave form feature database and the mapping effect of sample vector in two-dimensional space be as shown in Figure 3.
More preferably, in step 2, support vector machine (Support Vector Machine is called for short SVM) mode of learning can be used: first weight is arranged to each latitude value of sample vector; Then use some sample vector data tests and adjust the weight of each vector; Finally use SVM judge new feature vector should belong to which region (such as flatly to crouch, sit up, lie on the left side and right side sleeping in where class vibration data) or whether should to be rejected.
Step S3 represents correction marginal date.Here marginal date, refers to that certain signal is in the data of region and region intersection in four regions shown in Fig. 3.Because there will be some ambiguous heart shock wave form signals unavoidably in the machine learning stage, likely namely belong to heart shock wave form signals when lying low, heart shock wave form signals during sitting may be belonged to again, that just cannot be correct in machine-learning process be divided in correct Sample Storehouse, so need in the light of actual conditions to revise these data.Correction can adopt various ways, can directly give up, or adopts threshold discrimination, also can artificial cognition etc.In this example, adopt the mode of artificial cognition, the artificial composition participated in may be needed many in early stage, along with increasing of later stage Sample Storehouse, judge more and more accurate, the composition that people participates in just can reduce rapidly.
By step S1 ~ S3, namely complete the foundation of heartbeat shock wave form feature database.
In sum, the present invention sets up heartbeat shock wave form feature database by the mode of machine learning, and efficiency is high, and accuracy is good.And according to preferred mode, simply, calculated amount is little for the sample vector of waveform and proper vector.So the present invention effectively overcomes various shortcoming of the prior art and tool high industrial utilization.
Above-described embodiment is illustrative principle of the present invention and effect thereof only, but not for limiting the present invention.Any person skilled in the art scholar all without prejudice under spirit of the present invention and category, can modify above-described embodiment or changes.Therefore, such as have in art usually know the knowledgeable do not depart from complete under disclosed spirit and technological thought all equivalence modify or change, must be contained by claim of the present invention.
Claims (10)
1. set up a method for heartbeat shock wave form feature database, it is characterized in that, comprising:
Select and add master sample vector;
New feature vector is added by the mode of machine learning;
Revise marginal date.
2. the method setting up heartbeat shock wave form feature database according to claim 1, is characterized in that, described selection the step of adding master sample vector comprise:
People's heartbeat shockwave information is gathered by measuring bed vibrations;
Sample waveform is selected from the described heartbeat shockwave information collected;
For described sample waveform selects unique point;
Sample vector is generated according to the position relationship between described unique point or mathematical relation;
Described sample vector is added to described heartbeat shock wave form feature database.
3. the method setting up heartbeat shock wave form feature database according to claim 2, is characterized in that, the step that the described mode by machine learning adds new feature vector comprises:
The heartbeat surge waveform newly collected is converted into proper vector;
Judge the similarity of described proper vector and described sample vector, if described similarity is higher than pre-determined threshold, then described proper vector is added to described heartbeat shock wave form feature database; Otherwise give up described proper vector.
4. the method setting up heartbeat shock wave form feature database according to claim 2, is characterized in that, the described mode by machine learning is added in the step of new feature vector, uses SVM to add new feature vector.
5. the method setting up heartbeat shock wave form feature database according to claim 2, is characterized in that, the step generating sample vector according to the position relationship between described unique point or mathematical relation comprises:
Using first dimension values of the distance between described unique point as described sample vector;
Using second dimension values of the slope of described unique point line as described sample vector.
6. the method setting up heartbeat shock wave form feature database according to claim 3, is characterized in that, described similarity represents with the distance of described proper vector and described sample vector.
7. the method setting up heartbeat shock wave form feature database according to claim 5, is characterized in that, described sample vector is the multi-C vector being greater than dimension, and the third dimension angle value of described sample vector is described unique point horizontal stroke/ordinate variance; The third dimension angle value of described sample vector is described unique point horizontal stroke/ordinate quadratic sum.
8. the method setting up heartbeat shock wave form feature database according to claim 6, is characterized in that, described selection is also added in the step of master sample vector, and described sample vector at least comprises two.
9. the method setting up heartbeat shock wave form feature database according to claim 2, it is characterized in that, describedly to gather in the step of people heartbeat shockwave information by measuring bed vibrations, comprise gather people static flat sleeping in bed, staticly to sit up in bed, staticly to lie on the left side in bed, the vibration information of the heartbeat shockwave information of static right side when crouching in bed and people bed when getting in and out of bed.
10. according to the method setting up heartbeat shock wave form feature database in claim 1-9 described in any one, it is characterized in that, described selection is also added described in master sample vector sum and is added in the step of new feature vector by the mode of machine learning, also comprises and the identity information of measured and measured's action classification information being added in described heartbeat shock wave form feature database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310327012.0A CN104346369B (en) | 2013-07-30 | 2013-07-30 | A kind of method for establishing heartbeat shock wave form feature database |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310327012.0A CN104346369B (en) | 2013-07-30 | 2013-07-30 | A kind of method for establishing heartbeat shock wave form feature database |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104346369A true CN104346369A (en) | 2015-02-11 |
| CN104346369B CN104346369B (en) | 2018-03-23 |
Family
ID=52501992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310327012.0A Active CN104346369B (en) | 2013-07-30 | 2013-07-30 | A kind of method for establishing heartbeat shock wave form feature database |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104346369B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108542381A (en) * | 2018-04-11 | 2018-09-18 | 京东方科技集团股份有限公司 | A kind of data processing method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040181543A1 (en) * | 2002-12-23 | 2004-09-16 | Canon Kabushiki Kaisha | Method of using recommendations to visually create new views of data across heterogeneous sources |
| CN101216884A (en) * | 2007-12-29 | 2008-07-09 | 北京中星微电子有限公司 | A method and system for face authentication |
| CN102397064A (en) * | 2011-12-14 | 2012-04-04 | 中国航天员科研训练中心 | Continuous blood pressure measuring device |
| CN102929872A (en) * | 2011-08-08 | 2013-02-13 | 阿里巴巴集团控股有限公司 | Computer-implemented message filtering method, message filtering device and system |
-
2013
- 2013-07-30 CN CN201310327012.0A patent/CN104346369B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040181543A1 (en) * | 2002-12-23 | 2004-09-16 | Canon Kabushiki Kaisha | Method of using recommendations to visually create new views of data across heterogeneous sources |
| CN101216884A (en) * | 2007-12-29 | 2008-07-09 | 北京中星微电子有限公司 | A method and system for face authentication |
| CN102929872A (en) * | 2011-08-08 | 2013-02-13 | 阿里巴巴集团控股有限公司 | Computer-implemented message filtering method, message filtering device and system |
| CN102397064A (en) * | 2011-12-14 | 2012-04-04 | 中国航天员科研训练中心 | Continuous blood pressure measuring device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108542381A (en) * | 2018-04-11 | 2018-09-18 | 京东方科技集团股份有限公司 | A kind of data processing method and device |
| CN108542381B (en) * | 2018-04-11 | 2021-08-31 | 京东方科技集团股份有限公司 | Data processing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104346369B (en) | 2018-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Multi-model ensemble with rich spatial information for object detection | |
| Zhao et al. | The functional data analysis view of longitudinal data | |
| CN110619638A (en) | Multi-mode fusion significance detection method based on convolution block attention module | |
| CN108664893A (en) | A kind of method for detecting human face and storage medium | |
| CN111754395B (en) | A Robustness Evaluation Method of Brain Function Supernetwork Model | |
| Sun et al. | Effectiveness of cartogram for the representation of spatial data | |
| CN104318270A (en) | Land cover classification method based on MODIS time series data | |
| CN107133651A (en) | The functional magnetic resonance imaging data classification method of subgraph is differentiated based on super-network | |
| CN102867115A (en) | Farmland division method based on fuzzy c-means clustering | |
| CN109817339A (en) | Patient's group technology and device based on big data | |
| CN105608459A (en) | Commodity image segmentation method and commodity image segmentation device | |
| CN101696968A (en) | New method for monitoring heavy metal content in soil | |
| CN106650572A (en) | Method for assessing quality of fingerprint image | |
| CN110111885A (en) | Attribute forecast method, apparatus, computer equipment and computer readable storage medium | |
| CN102855490A (en) | Object-neural-network-oriented high-resolution remote-sensing image classifying method | |
| Gao et al. | A novel face feature descriptor using adaptively weighted extended LBP pyramid | |
| Calvet et al. | Unsupervised clustering and performance prediction of vortex wakes from bio-inspired propulsors | |
| CN104778341A (en) | Magnetic resonance coil combination coefficient calculation method, magnetic resonance imaging method and device thereof | |
| CN102841963B (en) | Method for detecting scale sensitivity of cellular automaton model | |
| CN104346369A (en) | Method of establishing heartbeat impulse wave form feature library | |
| CN103268423A (en) | A Method of Spatial Scale Selection for Multi-point Simulation of Geographical Phenomena | |
| KR20200005853A (en) | Method and System for People Count based on Deep Learning | |
| Yang et al. | RDHCformer: Fusing ResDCN and transformers for fetal head circumference automatic measurement in 2D ultrasound images | |
| CN108268478A (en) | A kind of unbalanced dataset feature selection approach and device based on ur-CAIM algorithms | |
| CN111222528A (en) | An Improved SSD Object Detection Algorithm with Region Enlargement Operation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |