CN108549900A - Tumble detection method for human body based on mobile device wearing position - Google Patents
Tumble detection method for human body based on mobile device wearing position Download PDFInfo
- Publication number
- CN108549900A CN108549900A CN201810188212.5A CN201810188212A CN108549900A CN 108549900 A CN108549900 A CN 108549900A CN 201810188212 A CN201810188212 A CN 201810188212A CN 108549900 A CN108549900 A CN 108549900A
- Authority
- CN
- China
- Prior art keywords
- wearing position
- mobile device
- value
- human body
- detection method
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1116—Determining posture transitions
- A61B5/1117—Fall detection
-
- 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)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Physiology (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The present invention provides a kind of tumble detection method for human body based on mobile device wearing position, including:The feature extracting method for using rotary mode component and attitude angle to merge first, calculates radius of turn, angular speed amplitude, attitude angle using accelerometer and gyro data and extracts feature, then classified the wearing position for obtaining mobile device;A kind of fall detection algorithm based on Time-Series analysis is then adaptively adjusted according to position.The mobile device wearing position discrimination of this method is 95.32%, can accurately distinguish the wearing position of user's mobile device;All it is optimal in the accuracy rate of different location, Time-Series analysis fall detection algorithm, 92% or more.
Description
Technical field
The invention belongs to the communications fields, and in particular to a kind of human body fall detection side based on mobile device wearing position
Method.
Background technology
Currently, having occurred a variety of wearable fall detection warning devices in the market.Those fall detections alarm dress
It sets and using being both needed to be worn on privileged site of body, such as wrist, loins etc..Since different wearing positions can be examined to falling
It surveys model and parameter impacts, and then influence the recognition effect of fall detection, would generally be required before user's use for the first time
Equipment is worn on designated position.For user during actual use, the wearing position of mobile phone is often in variation.Generally
In the case of, the elderly would generally switch the equipment such as smart mobile phone between three positions of body wearing:Wrist, loins, pocket.
Since there is tumble event randomness, the wearing position of mobile device also to have uncertainty, same fall detection algorithm is not
Recognition effect with position has larger difference.
Currently, the fall detection algorithm based on single wearing position more uses threshold detection method.Threshold test is calculated
Although method has many advantages, such as that design is simple, computing cost is small, poor for complicated situation adaptability, can not be in each wearing
Position all has good detection result.
Invention content
The invention discloses a kind of tumble detection method for human body based on mobile device wearing position.This method fully considers
The wearing position of the common mobile devices such as wrist, pocket, loins, and according to the tumble of wearing position self-adapting detecting human body
Situation makes the fall detection of different wearing positions identify and all reaches highest.
For achieving the above object, the present invention provides following technical scheme:
A kind of tumble detection method for human body based on mobile device wearing position, including:
(1) training sample is built:The 3-axis acceleration and angular velocity data of several users are acquired by motion sensor, and
Feature extraction is carried out to 3-axis acceleration and angular velocity data, obtains characteristic set (Xr,Xω,Xpitch,Xroll), respectively to every
A characteristics extraction characteristic component mean value, variance, intermediate value, kurtosis, the degree of bias, quarter back's number, constitute the corresponding spy of each characteristic value
Levy subset A, wherein XrIndicate radius of turn eigenmatrix, XωIndicate angular speed eigenmatrix, XpitchIndicate attitude angle Pitch
Eigenmatrix, XrollIndicate attitude angle Roll eigenmatrixes;
(2) using all character subset A as training sample, training Logistic regression models obtain wearing position
Disaggregated model;
(3) test sample is built using the identical method with step (1), using wearing position disaggregated model to test sample
It is predicted, determines the corresponding wearing position of test sample data;
(4) it is directed to the vector sum SMV of the corresponding 3-axis acceleration of each wearing position, the characteristic component for extracting SMV is maximum
Value, minimum value, mean value, range, variance, standard deviation, root mean square, signal amplitude area, quartile, absolute value, and screen spy
5 characteristic component, constitutive characteristic subset B before sign component score value ranking;
(5) using the corresponding all character subset B of each wearing position as training sample, svm classifier mould is respectively trained
Type obtains the corresponding fall detection model of each wearing position;
(6) after according to step (1) to 3-axis acceleration and the angular velocity data processing of acquisition, it is input to wearing position classification
Model obtains the wearing position of user's mobile device, and according to step (2) to the corresponding 3-axis acceleration number of the wearing position
After being handled, it is input to fall detection model corresponding with the wearing position, exports fall detection result.
Detection method provided by the invention has universality, and the wearing position of mobile device can be determined according to the data of acquisition
It sets, and corresponding fall detection model can be selected according to wearing position, to improve the accuracy for detecting whether to fall.
Preferably, in step (1), based on the time window of a length of 512 sampled points, according to 50% time slip-window pair
3-axis acceleration and angular velocity data carry out feature extraction.
Preferably, in step (2), in Logistic regression models, the logarithmic loss function of training pattern is:
Wherein, x inputs for sample,It is exported for model, θ is the model parameter of training pattern, and y is sample
Corresponding concrete class value,
With the minimum condition of convergence of loss function value, solver uses coordinate descent algorithm, along reference axis direction into
Row parameter updates, and regularization mode is L2 regularizations, and parameter renewal process is as follows:
A) initial parameter is chosen
B) it is iterated for currently available parameter, it is assumed that the parameter of the wheel of kth -1 has been found out,
The parameter renewal process of kth wheel is as follows:
…
C) iteration result of every wheel is obtained by above step, if θkRelative to θk‐1Vary less, then stop change
Otherwise in generation, repeats step b).
Preferably, in step (4), SMV segmentations is carried out to each acquisition sliding time window, are divided into 15 segments, to every
SMV data in segment carry out feature extraction.Further, feature extraction is carried out to SMV data using Filter filtration methods,
Obtain characteristic component.
Preferably, in step (5), in training SVM models, the loss function of training pattern is:
Wherein, θ, b are the parameter of Optimal Separating Hyperplane, yiFor the corresponding concrete class value of sample, kernel function f (xi) using high
This kernel function, C are the penalty coefficients of L2 regularizations.
Loss function value is minimum while the condition of convergence of training pattern requires to optimize.C is bigger, the energy of fit non-linear
Power is stronger, and gamma values are bigger, more insensitive to noise.Using Grid Search methods respectively to penalty coefficient C and Gaussian kernel letter
Gamma values in number optimize.
Preferably, the mobile device includes smart mobile phone, Intelligent bracelet, smartwatch, intelligent pendant, intelligent waistband.
Compared with prior art, the present invention have the advantage that for:
Using tumble detection method for human body provided by the invention, the discrimination to wearing position is 95.32%, Ke Yizhun
Really distinguish the wearing position of user's mobile device;In different location, the accuracy rate whether human body falls all is optimal,
92% or more.
Description of the drawings
Fig. 1 is hyperspin feature component and hyperspin feature component+attitude angle classification accuracy rate comparison diagram in embodiment;
Fig. 2 is tumble process SMV time window stepwise schematic views in embodiment;
Fig. 3 is the fall detection confusion matrix schematic diagram on each wearing position test set in embodiment;
Fig. 4 is human body fall detection flow chart in embodiment.
Specific implementation mode
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, with reference to the accompanying drawings and embodiments to this
Invention is described in further detail.It should be appreciated that the specific embodiments described herein are only used to explain the present invention,
Do not limit protection scope of the present invention.
In the present embodiment, the process of establishing of wearing position disaggregated model is:
The public data collection REALWORLD2016 created using Mannheim, Germany university Timo Sztyler professors et al.
Method is evaluated and tested.Data set includes that 15 users (age is between 19~45 years old) 3 motion sensors of wearing are engaged in 8
Sensing data recorded in item active procedure.8 Activity Types include:(A upstairs1), downstairs (A2), jump (A3), lie
Under (A4), stand (A5), sit down (A6), run/jog (A7), walk (A8).The position of wearable sensors includes wrist
(forearm), pocket (thigh), loins (waist).Other than jump action, each same action of tester's same position is surveyed
It is about 10 minutes to try duration.Each sensor is in data acquisition with the frequency collection 3-axis acceleration and angular speed of 50Hz
Data.The data set covers each anthropoid daily behavior action, and tumble situation also mostly occurs under above-mentioned action.
Experimental arrangement is write using Python, and wherein Logistic recurrence is realized based on Maximum-likelihood estimation.In reality
In testing, feature extraction is sampled according to 50% time slip-window based on the time window of a length of 512 sampled points.It is first
First from each sliding time window extraction characteristic set (Xr,Xω,Xpitch,Xroll), then extracted in table 1 from each feature
Character subset.Finally, the sample size of experimental data set is 37935, and the character subset quantity of every group of sample is 32.
1 radius of turn component of table and attitude angle character subset
In training process, data set is divided into 70% training set and 30% test set first.Again on training set
The phenomenon that over-fitting being avoided using 10 folding cross-validation methods.Experimental result from the result in Fig. 1 as shown in Figure 1, can be seen that:This
The rotary mode that embodiment is proposed adds the combination of attitude angle (rotation+attitude mode) feature that can obtain preferably
Effect, this method can obtain 95.32% cross validation accuracy rate.And traditional rotary mode component characterization that is based only upon accounts for
Excellent method only obtains 92.18% cross validation accuracy rate.Mobile device in data set is especially increased obviously not rotate
The data of situation, the addition of posture corner characteristics can play the ability for improving and integrally classifying.
The process of establishing of fall detection model is:
Using Filter filtration methods, the subset of three wearing positions is screened respectively.Character subset closes as shown in table 2 below.
2 fall detection common feature value of table
Table 3 is the characteristic value of the highest scoring filtered out according to feature marking value size.Each wearing position selects score
Highest five characteristic values, as the extraction character subset in the section of linear segmented.
3 each position feature selecting result of table
In the present embodiment, using to time window carry out stage extraction feature by the way of, purpose be by it is adjacent and become
The close acceleration value of change trend is divided in the same segmentation, and the variation tendency of acceleration value is dramatically different between each section.
In data preprocessing phase, a length of 6s of original time window, acceleration frequency acquisition is still 50Hz.Using base
Each sliding time window SMV is segmented in the improved Piecewise Linear Representation methods of PAA, as shown in Figure 2.Time window quilt
It is divided into 15 segments, each section carries out characteristic component extraction according to the characteristic component and Filter filtration methods that are determined in table 3.Signal width
It is worth area (SMA) and absolute value | Ai| it is calculated with initial 3-axis acceleration data, being averaged in every section is extracted according to segmentation method
Value.The characteristic component of extraction is trained model as training sample.
Specifically, 17907 groups of ADL data and 6110 groups of Fall data have been used in the present embodiment, have still drawn data set
It is divided into 70% training set and 30% test set, then showing for over-fitting is avoided using 10 folding cross-validation methods on training set
As being trained to support vector machines (SVM) disaggregated model, obtaining fall detection model.
In order to assess test effect, 3 parameters are defined:Accuracy rate (AR), recall rate (DR), false alarm rate, form are as follows.
Wherein p and q respectively represents the number of positive sample in data sample (Fall) and negative sample (ADL).Correspondingly, TP
The number that positive sample is correctly validated is represented, TN represents the number that negative sample is correctly identified as in negative sample, and FP, which is represented, bears sample
The number for positive sample is accidentally known in this.Therefore, AR is defined as the number that positive and negative sample standard deviation is correctly validated and accounts for all samples
Percentage, DR are defined as the probability that negative sample is correctly validated, and FAR then represents the probability that negative sample is reported by mistake.Confusion matrix
Us can be helped more clearly to obtain the value of AR, DR and FAR.Fig. 3 represents SVM training patterns and obscures square on test set
Battle array.
Table 4 gives the fall detection assessment result of each wearing position.In contrast experiment, the sequential fallen is not considered
Property, maximum value, minimum value, the average value etc. of SMV are extracted in a complete time window.
4 experimental result of table
The experimental results showed that fall detection model used by the present embodiment has stronger wearing position adaptability, move
Dynamic equipment is placed on any wearing site and achieves good fall detection effect.It can reach higher when being fixed on loins to fall
Discrimination is detected, accuracy rate can reach 96.48%.When mobile device is placed on hand or in pocket, although movement is set
It is standby to be constantly in shaking or rotary state, but extract suitable feature and can still obtain preferable classifying quality.As
Contrast experiment, the feature extracted in a complete time window is only capable of reflecting tumble process roughly, to each during tumble
Variation reflection between stage is less, thus classification accuracy is not so good as temporal analysis on the whole.The sequential used herein, which is fallen, to be examined
Method of determining and calculating selects most suitable training characteristics to different wearing positions, the fall detection discrimination of each position reached 92% with
On, there is larger practical value.
After above-mentioned two model foundation is good, the model is used to carry out the detailed process of human body fall detection as such as Fig. 4
It is shown:
First, after in the way of step in invention content (1) to the processing of the 3-axis acceleration and angular velocity data of acquisition,
It is input to wearing position disaggregated model, obtains the wearing position of user's mobile device;
Then, it after being handled the corresponding 3-axis acceleration data of the wearing position according to step (2), is input to and is somebody's turn to do
The corresponding fall detection model of wearing position exports fall detection result.
Finally, when testing result is to fall, alarm can be sent out, otherwise re-starts data acquisition and prediction.
Method provided in this embodiment considers the wearing position of the common mobile device such as wrist, pocket, loins, first
Wearing position is detected using the method that rotary mode component and attitude angle merge, then uses a kind of base in correspondingly wearing position
In the fall detection method of Time-Series analysis.This method is based on wearing position, and characteristic component is extracted using Filter Method for Feature Selection,
Correspondingly characteristic model is trained, so that the fall detection of different wearing positions is identified and all reaches highest.
Technical scheme of the present invention and advantageous effect is described in detail in above-described specific implementation mode, Ying Li
Solution is not intended to restrict the invention the foregoing is merely presently most preferred embodiment of the invention, all principle models in the present invention
Interior done any modification, supplementary, and equivalent replacement etc. are enclosed, should all be included in the protection scope of the present invention.
Claims (7)
1. a kind of tumble detection method for human body based on mobile device wearing position, including:
(1) training sample is built:The 3-axis acceleration and angular velocity data of several users are acquired by motion sensor, and to three
Axle acceleration and angular velocity data carry out feature extraction, obtain characteristic set (Xr,Xω,Xpitch,Xroll), respectively to each feature
Value extraction characteristic component mean value, variance, intermediate value, kurtosis, the degree of bias, quarter back's number, constitute the corresponding character subset of each characteristic value
A, wherein XrIndicate radius of turn eigenmatrix, XωIndicate angular speed eigenmatrix, XpitchIndicate attitude angle Pitch feature squares
Battle array, XrollIndicate attitude angle Roll eigenmatrixes;
(2) using all character subset A as training sample, training Logistic regression models obtain wearing position classification
Model;
(3) test sample is built using the identical method with step (1), test sample is carried out using wearing position disaggregated model
Prediction, determines the corresponding wearing position of test sample data;
(4) it is directed to the vector sum SMV of the corresponding 3-axis acceleration of each wearing position, extracts the characteristic component maximum value, most of SMV
Small value, mean value, range, variance, standard deviation, root mean square, signal amplitude area, quartile, absolute value, and screen characteristic component
5 characteristic component before score value ranking, constitutive characteristic subset B;
(5) using the corresponding all character subset B of each wearing position as training sample, svm classifier model is respectively trained, obtains
Obtain the corresponding fall detection model of each wearing position;
(6) after according to step (1) to 3-axis acceleration and the angular velocity data processing of acquisition, it is input to wearing position classification mould
Type, obtain user's mobile device wearing position, and according to step (2) to the corresponding 3-axis acceleration data of the wearing position into
After row processing, it is input to fall detection model corresponding with the wearing position, exports fall detection result.
2. the tumble detection method for human body as described in claim 1 based on mobile device wearing position, which is characterized in that step
(1) in, based on the time window of a length of 512 sampled points, according to 50% time slip-window to 3-axis acceleration and angular speed
Data carry out feature extraction.
3. the tumble detection method for human body as described in claim 1 based on mobile device wearing position, which is characterized in that step
(2) in, in Logistic regression models, the logarithmic loss function of training pattern is:
Wherein, x inputs for sample,It is exported for model, θ is the model parameter of training pattern, and y corresponds to for sample
Concrete class value,
With the minimum condition of convergence of loss function value, solver uses coordinate descent algorithm, is joined along the direction of reference axis
Number update, regularization mode are L2 regularizations, and parameter renewal process is as follows:
A) initial parameter is chosen
B) it is iterated for currently available parameter, it is assumed that the parameter of the wheel of kth -1 has been found out,
The parameter renewal process of kth wheel is as follows:
C) iteration result of every wheel is obtained by above step, if θkRelative to θk‐1Vary less, then stop iteration, it is no
Then, step b) is repeated.
4. the tumble detection method for human body as described in claim 1 based on mobile device wearing position, which is characterized in that step
(4) in, SMV segmentations is carried out to each acquisition sliding time window, are divided into 15 segments, the SMV data in every segment are carried out
Feature extraction.
5. the tumble detection method for human body as claimed in claim 4 based on mobile device wearing position, which is characterized in that use
Filter filtration methods carry out feature extraction to SMV data, obtain characteristic component.
6. the tumble detection method for human body as described in claim 1 based on mobile device wearing position, which is characterized in that step
(5) in, in training SVM models, the loss function of training pattern is:
Wherein, θ, b are the parameter of Optimal Separating Hyperplane, yiFor the corresponding concrete class value of sample, kernel function f (xi) use Gaussian kernel
Function, C are the penalty coefficients of L2 regularizations.
The condition of convergence of training pattern is while requiring to optimize, and loss function value is minimum, and C is bigger, the energy of fit non-linear
Power is stronger, and gamma values are bigger, more insensitive to noise, using Grid Search methods respectively to penalty coefficient C and Gaussian kernel letter
Gamma values in number optimize.
7. the tumble detection method for human body as described in claim 1 based on mobile device wearing position, which is characterized in that described
Mobile device includes smart mobile phone, Intelligent bracelet, smartwatch, intelligent pendant, intelligent waistband.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810188212.5A CN108549900A (en) | 2018-03-07 | 2018-03-07 | Tumble detection method for human body based on mobile device wearing position |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810188212.5A CN108549900A (en) | 2018-03-07 | 2018-03-07 | Tumble detection method for human body based on mobile device wearing position |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108549900A true CN108549900A (en) | 2018-09-18 |
Family
ID=63516360
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810188212.5A Pending CN108549900A (en) | 2018-03-07 | 2018-03-07 | Tumble detection method for human body based on mobile device wearing position |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108549900A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110222708A (en) * | 2019-04-29 | 2019-09-10 | 中国科学院计算技术研究所 | A kind of fall detection method and system based on Integrated Decision tree |
CN110245744A (en) * | 2019-05-29 | 2019-09-17 | 金华诺普视信息技术研究所有限公司 | It is a kind of that detection method is fallen down based on multilayer perceptron |
CN110414590A (en) * | 2019-07-24 | 2019-11-05 | 重庆大学 | Physical activity recognition methods based on Intelligent mobile equipment and convolutional neural networks |
CN110659595A (en) * | 2019-09-10 | 2020-01-07 | 电子科技大学 | Tumble type and injury part detection method based on feature classification |
CN111158494A (en) * | 2020-01-15 | 2020-05-15 | 山东师范大学 | Posture correction device and posture correction method |
CN111296994A (en) * | 2019-12-20 | 2020-06-19 | 石狮市森科智能科技有限公司 | Intelligent gesture interaction control system |
CN112580403A (en) * | 2019-09-29 | 2021-03-30 | 北京信息科技大学 | Time-frequency feature extraction method for fall detection |
CN112699744A (en) * | 2020-12-16 | 2021-04-23 | 南开大学 | Fall posture classification identification method and device and wearable device |
CN116070105A (en) * | 2023-03-17 | 2023-05-05 | 湖北工业大学 | Combined beam damage identification method and system based on wavelet transformation and residual error network |
FR3138843A1 (en) * | 2022-08-10 | 2024-02-16 | Nov'in | Device for monitoring user activity |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060284979A1 (en) * | 2005-06-09 | 2006-12-21 | Sony Corporation | Activity recognition apparatus, method and program |
CN103308069A (en) * | 2013-06-04 | 2013-09-18 | 电子科技大学 | Falling-down detection device and method |
CN103377541A (en) * | 2013-07-16 | 2013-10-30 | 中国科学院深圳先进技术研究院 | Human body fall-down prevention early warning and intervening method and system |
CN103544485A (en) * | 2013-10-30 | 2014-01-29 | 无锡赛思汇智科技有限公司 | Driver recognition method and device based on intelligent terminal |
CN103968827A (en) * | 2014-04-09 | 2014-08-06 | 北京信息科技大学 | Wearable human body gait detection self-localization method |
CN104217107A (en) * | 2014-08-27 | 2014-12-17 | 华南理工大学 | Method for detecting tumbling state of humanoid robot based on multi-sensor information |
CN104243656A (en) * | 2014-10-10 | 2014-12-24 | 北京大学工学院南京研究院 | Auto-dialing distress method used after user falling detected by smart phone |
CN104268577A (en) * | 2014-06-27 | 2015-01-07 | 大连理工大学 | Human body behavior identification method based on inertial sensor |
CN105590409A (en) * | 2016-02-26 | 2016-05-18 | 江苏大学 | Human body tumble detection method and human body tumble detection system based on big data |
CN105590408A (en) * | 2016-02-06 | 2016-05-18 | 高强 | Human body falling detection method and protection device |
US20170039045A1 (en) * | 2015-08-06 | 2017-02-09 | Avishai Abrahami | Cognitive state alteration system integrating multiple feedback technologies |
CN207008661U (en) * | 2017-07-25 | 2018-02-13 | 江苏淼升轨道科技有限公司 | The goods that a kind of WSN is combined with RFID is in way attitude monitoring system |
-
2018
- 2018-03-07 CN CN201810188212.5A patent/CN108549900A/en active Pending
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060284979A1 (en) * | 2005-06-09 | 2006-12-21 | Sony Corporation | Activity recognition apparatus, method and program |
CN103308069A (en) * | 2013-06-04 | 2013-09-18 | 电子科技大学 | Falling-down detection device and method |
CN103377541A (en) * | 2013-07-16 | 2013-10-30 | 中国科学院深圳先进技术研究院 | Human body fall-down prevention early warning and intervening method and system |
CN103544485A (en) * | 2013-10-30 | 2014-01-29 | 无锡赛思汇智科技有限公司 | Driver recognition method and device based on intelligent terminal |
CN103968827A (en) * | 2014-04-09 | 2014-08-06 | 北京信息科技大学 | Wearable human body gait detection self-localization method |
CN104268577A (en) * | 2014-06-27 | 2015-01-07 | 大连理工大学 | Human body behavior identification method based on inertial sensor |
CN104217107A (en) * | 2014-08-27 | 2014-12-17 | 华南理工大学 | Method for detecting tumbling state of humanoid robot based on multi-sensor information |
CN104243656A (en) * | 2014-10-10 | 2014-12-24 | 北京大学工学院南京研究院 | Auto-dialing distress method used after user falling detected by smart phone |
US20170039045A1 (en) * | 2015-08-06 | 2017-02-09 | Avishai Abrahami | Cognitive state alteration system integrating multiple feedback technologies |
CN105590408A (en) * | 2016-02-06 | 2016-05-18 | 高强 | Human body falling detection method and protection device |
CN105590409A (en) * | 2016-02-26 | 2016-05-18 | 江苏大学 | Human body tumble detection method and human body tumble detection system based on big data |
CN207008661U (en) * | 2017-07-25 | 2018-02-13 | 江苏淼升轨道科技有限公司 | The goods that a kind of WSN is combined with RFID is in way attitude monitoring system |
Non-Patent Citations (2)
Title |
---|
SZTYLER, T等: "n-body localization of wearable devices:An investigation of position-aware activity recognition", 《2016 IEEE INTERNATIONAL CONFERENCE ON PERVASIVE COMPUTING AND COMMUNICATIONS》 * |
时岳等: "基于旋转模式的移动设备佩戴位置识别方法", 《软件学报》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110222708A (en) * | 2019-04-29 | 2019-09-10 | 中国科学院计算技术研究所 | A kind of fall detection method and system based on Integrated Decision tree |
CN110245744A (en) * | 2019-05-29 | 2019-09-17 | 金华诺普视信息技术研究所有限公司 | It is a kind of that detection method is fallen down based on multilayer perceptron |
CN110414590A (en) * | 2019-07-24 | 2019-11-05 | 重庆大学 | Physical activity recognition methods based on Intelligent mobile equipment and convolutional neural networks |
CN110659595A (en) * | 2019-09-10 | 2020-01-07 | 电子科技大学 | Tumble type and injury part detection method based on feature classification |
CN112580403A (en) * | 2019-09-29 | 2021-03-30 | 北京信息科技大学 | Time-frequency feature extraction method for fall detection |
CN111296994A (en) * | 2019-12-20 | 2020-06-19 | 石狮市森科智能科技有限公司 | Intelligent gesture interaction control system |
CN111158494A (en) * | 2020-01-15 | 2020-05-15 | 山东师范大学 | Posture correction device and posture correction method |
CN111158494B (en) * | 2020-01-15 | 2023-10-03 | 山东师范大学 | Posture correction device and posture correction method |
CN112699744A (en) * | 2020-12-16 | 2021-04-23 | 南开大学 | Fall posture classification identification method and device and wearable device |
FR3138843A1 (en) * | 2022-08-10 | 2024-02-16 | Nov'in | Device for monitoring user activity |
CN116070105A (en) * | 2023-03-17 | 2023-05-05 | 湖北工业大学 | Combined beam damage identification method and system based on wavelet transformation and residual error network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108549900A (en) | Tumble detection method for human body based on mobile device wearing position | |
CN107220617A (en) | Human body attitude identifying system and method | |
Wang et al. | Walking pattern classification and walking distance estimation algorithms using gait phase information | |
CN107316067B (en) | A kind of aerial hand-written character recognition method based on inertial sensor | |
CN102302370B (en) | Method and device for detecting tumbling | |
CN110245718A (en) | A kind of Human bodys' response method based on joint time-domain and frequency-domain feature | |
CN105877757A (en) | Multi-sensor integrated human motion posture capturing and recognizing device | |
CN110221699B (en) | Eye movement behavior identification method of front-facing camera video source | |
CN106981174A (en) | A kind of Falls Among Old People detection method based on smart mobile phone | |
CN108021888B (en) | Fall detection method | |
CN110113116B (en) | Human behavior identification method based on WIFI channel information | |
CN108196668B (en) | Portable gesture recognition system and method | |
CN108629170A (en) | Personal identification method and corresponding device, mobile terminal | |
CN106910314A (en) | A kind of personalized fall detection method based on the bodily form | |
CN107169334B (en) | The user authen method based on straight punch motion detection for hand wearable device | |
CN112464738B (en) | Improved naive Bayes algorithm user behavior identification method based on mobile phone sensor | |
CN107122711A (en) | A kind of night vision video gait recognition method based on angle radial transformation and barycenter | |
KR20130073361A (en) | Apparatus and method for classifing pattern of electromyogram signals | |
Jiang et al. | Development of a real-time hand gesture recognition wristband based on sEMG and IMU sensing | |
CN111582361A (en) | Human behavior recognition method based on inertial sensor | |
Ge et al. | Detecting Falls Using Accelerometers by Adaptive Thresholds in Mobile Devices. | |
CN105551191B (en) | A kind of fall detection method | |
CN114881079A (en) | Human body movement intention abnormity detection method and system for wearable sensor | |
Schmid et al. | SVM versus MAP on accelerometer data to distinguish among locomotor activities executed at different speeds | |
Lee et al. | Touchless hand gesture UI with instantaneous responses |
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: 20180918 |
|
RJ01 | Rejection of invention patent application after publication |