CN103211599A - Method and device for monitoring tumble - Google Patents
Method and device for monitoring tumble Download PDFInfo
- Publication number
- CN103211599A CN103211599A CN201310174295XA CN201310174295A CN103211599A CN 103211599 A CN103211599 A CN 103211599A CN 201310174295X A CN201310174295X A CN 201310174295XA CN 201310174295 A CN201310174295 A CN 201310174295A CN 103211599 A CN103211599 A CN 103211599A
- Authority
- CN
- China
- Prior art keywords
- human body
- falling
- tumble
- fallen
- inclination angle
- 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
Images
Landscapes
- Emergency Alarm Devices (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses a method and a device for monitoring tumble. The method for monitoring tumble comprises: utilizing a mode recognition method, creating a human body tumble behavior recognition model based on a support vector machine to screen standard feature vectors capable of serving as human body tumble judge, and judging the tumble state of the human body through the joint judge of the mode recognition of the support vector machine and the inclination angle threshold. The device is composed of an acceleration sensor, a microcontroller, a GPS (Global Position System) module, a GSM (Global System for Mobile Communications) module, and the like. The method for monitoring tumble is used for monitoring the acceleration and inclination angle data in a tumble process in real time, correctly recognizing the human body tumble state and timely sending the tumble information and direction to family members of the user and nurses of a community hospital to rescue the user in time. According to the invention, tumble detection is more reliable, tumble judging accuracy is improved, individual difference error is effectively reduced and recognition capability of different tumble categories is greatly improved.
Description
Technical field
The invention belongs to automatization's detection range, especially relate to method and device that a kind of monitoring is fallen.
Background technology
Falling is the common incident of injury of old people, be one of elderly population permanent disability, anergy and main causes of death, Health and Living to the old people threatens very big, bring huge burden and loss for family and society, being the important health care issues of old people, also is a social problem that should pay attention to.
Falling is the common incident of injury of old people, be one of elderly population permanent disability, anergy and main causes of death, Health and Living to the old people threatens very big, bring huge burden and loss for family and society, being the important health care issues of old people, also is a social problem that should pay attention to.
The old people is because handicapped, and often suffers from diseases such as hypertension, apoplexy, heart disease, and the incidence rate height of therefore falling, consequence are serious, become the first injury of old people at the advanced age cause of the death.Along with being on the rise of China's aged tendency of population, life and the operating pressure of younger generation under few sub-ization situation strengthens, be busy with work, the elderly's daily life shortage is effectively looked after and is accompanied and attended to, make the old people when falling, can not in time obtain relief, the quality of life and the life security that seriously jeopardize the old people.Therefore the fall behavior of old people in life monitored, when detecting the old man and fall down, can in time report to the police and seek help, so that in time sue and labour, to improving the life of elderly person quality and protection old people physical and mental health has great significance.
Present domestic great majority still just rest on the passive aspect of medical treatment nurse at old people's technical research.And at western developed country, the particularly U.S., Japan.In recent years, have many universities and research institution under the support of government and society fall down carried out aspect the detection technique hommization, to improve a large amount of research work that old man's quality of the life is a target.The method of falling for monitoring at present mainly contains three class basic skills: (1) is based on the analysis of video image, the real time kinematics of object is monitored by photographic head, its weak point is the personal secrets that can not guarantee the user, can only install and use at fixed-site, leave the camera head monitor scope and just can't work.(2) based on the analysis of sound signal, the incident of falling is partly judged by the frequency that the analysis impact causes vibrating, but it installs more complicated, and fund input is also bigger.(3) device based on Wearable detects, mainly contain two kinds: a kind of is to wear the fall variation at brief acceleration and inclination angle of accelerometer and inclinometer measurement by waist, again according to corresponding experiment with calculate and set threshold parameter and set to judge whether human body falls, but because the inclinometer asynchronism that algorithm is carried out when having work with accelerometer is easy to generate reports by mistake or fail to report.Another kind is that strain transducer is measured the vola stress or acceleration transducer is measured the acceleration of foot by installing at sole, realize falling down detection according to support vector machine, neutral net, scheduling algorithm again, because the signal detection aspect that this method monitoring is fallen exists some defectives, thereby has limited these practical application effects of very outstanding algorithm in theory.
Based on the recognizer of threshold value is to fall at present will be the research method of classics in the identification research.The algorithm design of the value of explaining method is more directly perceived, also be easier to comparatively speaking aspect the realization of algorithm, so this method is a most general present research method, its deficiency is that the setting that threshold value influences very big to recognition result, obtaining of threshold value is based upon on the experience or the basis of experimental data often, if usefulness be empirical value just be not easy to learn threshold value accuracy how, the people's that the fuzzy diagnosis of empirical value method causes easily individual variation error, the model of cognition complexity is higher.
Summary of the invention
In view of this, for solving the problems of the technologies described above, originally the method and the device that provide a kind of monitoring to fall, based on algorithm of support vector machine, can solve detection and differentiation problem that the old people falls down action process efficiently, reliably, can identify the behavior of falling by kinestate data and the measurement of dip angle and the Treatment Analysis of human bodies such as simple acceleration.
The method that a kind of monitoring that provides of the present invention is fallen comprises the following steps:
1) makes up based on the human body of the support vector machine behavior model of cognition of falling;
2) the kinestate data of real-time human body by the behavior model of cognition of falling based on the human body of support vector machine, judge whether human body falls.
3) warning after judgement is fallen
Described structure further comprises step based on the human body of the support vector machine behavior model of cognition of falling:
1) construction feature sample: gather the kinestate data of the different behavioral pattern of human body in the setting-up time, carry out forming kinestate sample characteristics vector after the pretreatment construction feature vector storehouse;
2) training classifier: according to based on the support vector machine principle, be that the feature samples that the kinestate data of pattern constitute is trained with the human body different rows respectively, obtain being used for the fall supporting vector machine model of state of human body;
3) make up the model of cognition of falling: from the training characteristics sample, extract the standard feature vector that is used to discern the behavior of falling, make up the template base of falling.
Preferably, described kinestate data are acceleration information.
Preferably, whether described judgement human body is fallen and is further comprised the judgement of whether the trunk inclination angle being satisfied the inclination angle setting value.
Preferably, whether described judgement human body is fallen and is further comprised the persistent period that inclination angle setting value is satisfied at the trunk inclination angle and whether reach the judgement of setting the persistent period.
The present invention further provides the device that a kind of monitoring is fallen, comprise microprocessor module and connected sensor assembly, audible alarm module, input button; Described sensor assembly is used for the kinestate data of human body; Described microprocessor module is used to receive described kinestate data, and data analysis is handled, and judges whether human body falls; Described audible alarm module is used to send alarm sound; Described input button is used for setting the mode of operation of microprocessor module to microprocessor module input external command and data; Described to the data analysis processing, whether the judgement human body is fallen and is comprised:
1) makes up based on the human body of the support vector machine behavior model of cognition of falling;
2) receive described kinestate data in real time,, judge whether human body falls by the behavior model of cognition of falling based on the human body of support vector machine.
Described structure can further comprise based on the human body of the support vector machine behavior model of cognition of falling:
1) construction feature sample: gather the kinestate data of the different behavioral pattern of human body in the setting-up time, carry out forming kinestate sample characteristics vector after the pretreatment construction feature vector storehouse;
2) training classifier: according to based on the support vector machine principle, be that the feature samples that the kinestate data of pattern constitute is trained with the human body different rows respectively, obtain being used for the fall supporting vector machine model of state of human body;
3) make up the model of cognition of falling: from the training characteristics sample, extract the standard feature vector that is used to discern the behavior of falling, make up the template base of falling.
Preferably, described sensor assembly is an acceleration transducer, described kinestate data are acceleration information, whether described judgement human body falls also comprises whether the trunk inclination angle being satisfied the judgement of inclination angle setting value and whether the persistent period that the inclination angle setting value is satisfied at the trunk inclination angle is reached the judgement of setting the persistent period, described trunk inclination angle is by obtaining described acceleration information computing.
The device that described a kind of monitoring is fallen also can further comprise GPS module, the wireless communication module that is connected with described microprocessor module; Described GPS module is used to measure the residing position of human body, and sends position data to microprocessor module; Described wireless communication module is used to receive the instruction and data that microprocessor module sends, and outwards reports information and orientation that human body is fallen.
Preferably, described wireless communication module is a gsm module.
The present invention is by adopting the method for pattern recognition based on support vector machine, the various interference of filtering effectively reduce the dimension of feature space, thereby significantly improve the accuracy and the reliability of fall detection; The present invention simultaneously has very high cost performance, uses the present invention can significantly reduce product cost, uses the sensor processor of low-power consumption to produce considerable energy-saving benefit, little convenient use that be easy to wear in light weight of this module volume.
Beneficial effect of the present invention:
1. realize utilizing human motion state data such as acceleration to carry out fall detection by support vector, the judgement of uniting of setting value is satisfied at its pattern recognition and inclination angle, makes that detection is more reliable, misrepresents deliberately with rate of false alarm low, speed is fast, has greatly improved and has judged the accuracy of falling.
2. the recognizer of falling that realizes of the thinking of utilization statistical model identification can reduce the people's that the fuzzy diagnosis of empirical value method causes individual variation error effectively.
3. by the method for pattern recognition, can come out the digitized feature description of various features that human body is fallen, reflect the feature of the process of falling preferably, greatly improve the fall identification ability of type of difference.
4. when for human body the comparatively complicated behavior of falling taking place, also good identification effect can be arranged.Acceleration digitized wave forms characteristic storage when falling is convenient to system optimization, and recognizer is easy to self study and renewal simultaneously.
Description of drawings
Fig. 1, the method step sketch map that the monitoring that the present invention proposes is fallen;
Fig. 2, the apparatus structure sketch map that the monitoring that the present invention proposes is fallen;
Fig. 3 makes up based on the human body of the support vector machine behavior model of cognition algorithm flow chart of falling
Fig. 4, the acceleration change curve chart in the process of falling;
Fig. 5, six kinds of type sketch maps of typically falling;
Fig. 6, the fall detection flow chart
Fig. 7, the fall detection device that the present invention proposes uses the wearing position sketch map
The specific embodiment
The invention will be further described below in conjunction with the drawings and specific embodiments:
At the human body differentiation problem of falling,, adopt support vector machine method to set up the fall detection model according to the thought of pattern recognition.At first, take all factors into consideration fall many-sided factor in the process of human body, adopt principal component analytical method to finish and depart from the angle of inclination Feature Extraction of vertical direction last trunk resultant acceleration, to set up fall detection category of model device by the excess syndrome method then and filter out and can be used as the characteristic vector that human body is fallen and judged, determine last trunk inclination angle threshold values in the falling over of human body action process by a large amount of experimental datas again.When a behavior has better similarity with the mode class of falling, just can or be divided into the mode class of falling with this behavior.The measurement index that each mode class is different from other mode class is the standard feature vector.
The behavior of falling is a kind of in people's numerous behaviors that may take place in life, self exclusive characteristic is arranged, as going downstairs, bend over, run Walk or the like tangible difference is arranged with other behavior, and these differences can be crossed some digitized feature descriptions and come out.The characteristic vector of digitized feature compositional model, the eigen vector difference of different mode can be as the fall distinguishing rule of pattern and other behavioral pattern of differentiation.Thus can be by selecting and extract the method for suitable feature value composition characteristic vector, will fall and other behavior makes a distinction.
As Fig. 1 and shown in Figure 6, the embodiment of the method that a kind of monitoring that provides of the present invention is fallen comprises the following steps:
(1) make up based on the human body of the support vector machine behavior model of cognition of falling, as shown in Figure 3, comprise step:
1) construction feature sample: gather the kinestate data of the different behavioral pattern of human body in the setting-up time, carry out forming kinestate sample characteristics vector after the pretreatment construction feature vector storehouse; Preferably, the kinestate data are acceleration information.
The different behavioral pattern of above-mentioned human body comprises the state of human body normal activity and the Status Type that human body is fallen.
In the human motion process, the different coordinate systems constantly of acceleration transducer change along with human posture's change in order to reduce computational complexity, eliminate the coordinate Mapping relation of different acceleration output valves constantly, introduce resultant acceleration AA (t)
A in the formula (1)
X, t, a
Y, t, a
Z, tRepresent the X that acceleration transducer collects respectively, Y, the acceleration on Z three direction of principal axis.
For supporting vector machine model is trained, gather the sample data of two groups of human motion states, comprise fall group and the non-group of falling (being normal group).As shown in Figure 5, the group of falling comprises that forward direction is fallen, the back to fall, falling in the left side, falls in the right side, face upward to lie and fall and the prostrate six kinds of types of falling of falling, Fig. 4 has shown the acceleration plots of six kinds of types of falling.The non-group of falling is mainly used in and comprises in the daily behavior that similar behavior of falling is such as going downstairs, be seated-lie down, stand-sit down, stand-the, Pao Walk that squats down, six kinds of typical normal behaviour types such as bend over.With 100Hz sample frequency (being that per second is gathered 100 groups of 3-axis acceleration values) respectively to six kinds of typical normal behaviour types with six kinds of type acceleration informations of falling are gathered and calculate corresponding resultant acceleration AA (t) by formula (1), through laboratory observation, this device judges that the human body time of the act of falling was about about 4 seconds, therefore, select to gather each 4 seconds interior sample datas of setting-up time, i.e. every type of each 400 sample data of six kinds of normal behaviour types and six kinds of types of falling.Choose that forward direction is fallen, the back to fall, falling in the left side, falls in the right side, face upward to lie and fall and the sample of the prostrate six kinds of types of falling of falling, the sample of six kinds of typical normal behaviour types such as going downstairs, be seated-lie down, stand-sit down, stand-the, Pao Walk that squats down, bend over constitutes sextuple characteristic vector respectively:
x
i=(AA
1, AA
2, AA
3, AA
4, AA
5, AA
6), x
iMiddle subscript i represents sample sequence, i=1 ..., 400.
Use w
kThe different conditions type of expression human body behavioral activity, w
kIn k=+1 or-1, w
+ 1The expression human body Status Type of falling, w
-1Expression human body normal behaviour active state type, the sample of the type of falling is positive sample sample, setting its sample class is y
i=w
+ 1, the sample of normal behaviour type is a negative sample, setting its sample class is y
i=w
-1
Form 400 characteristic vectors for every type, by the characteristic vector composition characteristic vector storehouse of all acquisitions.
2) training classifier: according to based on the support vector machine principle, be that the feature samples that the kinestate data of pattern constitute is trained with the human body different rows respectively, obtain being used for the fall support vector machine detector of state of human body;
According to based on the support vector machine principle, 400 movable with the human body normal behaviour respectively and the two states type of falling is respectively formed characteristic vectors, the composing training feature samples is trained, find the solution the double optimization problem, promptly can obtain being used for the human body Status Type w that falls
+ 1The support vector machine detector.
The identification problem that human body is fallen is that typical two classification problems are promptly judged and fallen and the non-state of falling, and for two classification problems of linear separability, obtaining sample set is (x
i, y
i), x ∈ R
6, i=1,2 ..., 400, y
iDesired output y for correspondence
i∈ { w
+ 1, w
-1.Have a hyperplane that this two class is separated fully, such hyperplane can be described as:
(w.x
i)+b=0 (2)
According to the support vector machine principle, utilize w
+ 1And w
-1The training characteristics sample training of two states is found the solution the double optimization problem, can be found the solution by Lagrange (Lagrange) method for the double optimization problem.Introduce the Lagrange multiplier, finding the solution of this problem can be converted into and find the solution its dual problem:
S.t.0≤a
i≤ C and
I=1 ..., N
The decision function that obtains the input data category is:
X in the formula
i, x
jIt is the characteristic vector of training sample; a
i, a
jBe and x
i, x
jCorresponding Lagrange multiplier, b is a constant, and C is the parameter that is used to represent to the punishment degree of classification error, and the C value 1 here, K (x
i, x
j) be the kernel function of nonlinear mapping correspondence, preferred, this kernel function is got radially base (RBF) function:
σ=0.5 of RBF preferably, is set.
3) make up the model of cognition of falling: from the training characteristics sample, extract the standard feature vector that is used to discern the behavior of falling, make up the template base of falling.
After support vector machine trained, can calculate Lagrange multiplier (Lagrangian coefficient) a of each training sample characteristic vector
i, corresponding Lagrange multiplier a
iSatisfy 0＜a
iThe sample characteristics vector of＜C and f (x) 〉=0 is known as support vector, and expression identifying object (human body behavior state) is in w
+ 1If the state state of promptly falling is Lagrange multiplier a
ia
iSatisfy 0＜a
i＜C and f (x)＜0, expression identifying object (human body behavior state) is in w
-1State is the normal behaviour kinestate.In view of the above, can from numerous characteristic vectors, extract the standard feature vector of differentiating as fall detection, form the template base of falling as training sample.
(2) the kinestate data of real-time human body by the behavior model of cognition of falling based on the human body of support vector machine, judge whether human body falls.
By sensor assembly (acceleration transducer), gather 4 seconds human motion state data in the setting-up time in real time with 100Hz sample frequency (being that per second is gathered 100 groups of 3-axis acceleration values), it is acceleration information, and calculate corresponding resultant acceleration AA (t) by formula (1), obtain 400 acceleration detection sample datas, composition characteristic vector sample.
Standard feature vector in the template base of falling is exactly the support vector machine that determines in the training process, the support vector machine decision making process can be regarded a kind of similarity comparison procedure as, series of standards characteristic vector in the characteristic vector sample and the template base of falling is compared, can identify the affiliated Status Type of characteristic vector sample, the similarity measurement of employing is exactly the RBF kernel function.
With the characteristic vector sample that real-time human body kinestate data are obtained, relatively vectorial with the standard feature of the template base of falling, by RBF kernel function similarity measurement, judge whether to belong to the characteristic vector of type of falling, thereby judge whether human body falls.
Preferably, whether described judgement human body is fallen and is further comprised the judgement of whether the trunk inclination angle being satisfied the inclination angle setting value.After human body is fallen, a tangible inclination angle takes place and changes in the torso portion of human body, what the measurement at inclination angle was adopted is that 360 ° of full-shapes are measured, main advantage can distinguished at each quadrant, and can measure by 360 ° of full-shapes, change and can both detect for the inclination angle of any direction of human body, be at the acceleration relation formula of plane camber angle θ and axle:
θ is the angle of two between centers in the formula, A
X, A
YBe respectively the static acceleration of X-direction and the static acceleration of Y direction.
In the process of falling, the trunk inclination angle with relatively have a bigger variable quantity with the body obliquity of original state, when the characteristic vector sample that obtains when real-time detection is identified as and falls type, calculate body obliquity variation delta θ by formula (2), with the inclination angle setting value that presets, i.e. inclination angle threshold values θ
LimitDo contrast, finally judge whether to fall, as Δ θ 〉=θ
LimitThe time, can judge that then human body falls.Described inclination angle threshold values θ
LimitCan draw by a large amount of experimental datas.
Preferably, whether described judgement human body is fallen and is further comprised the persistent period that inclination angle setting value is satisfied at the trunk inclination angle and whether reach the judgement of setting the persistent period.
In order to eliminate the wrong report that can stand rapidly after human body is fallen, the characteristic vector sample that obtains when real-time detection is identified as the type of falling, and body obliquity variation delta θ 〉=θ
LimitThe time, begin that the time of this kind state continuance is carried out timing and measure, if through body obliquity variation delta θ 〉=θ after 4 seconds setting persistent period
LimitCondition also satisfies, and judges finally that then the people falls.
(3) warning after judgement is fallen.After the final decision human body is fallen, by various existing type of alarms, the information that report is fallen.
As shown in Figure 2, the invention provides the device that a kind of monitoring is fallen, comprise microprocessor module and coupled sensor assembly, GPS module, audible alarm module, wireless communication module, input button; Sensor assembly is used for the kinestate data of human body; Microprocessor module is equipped with and is used to monitor the software of falling, and can receive described kinestate data, and data analysis is handled, and judges whether human body falls; The audible alarm module is used to send alarm sound; The input button is used for setting the mode of operation of microprocessor module to microprocessor module input external command and data; The GPS module is used to measure the residing position of human body, and sends position data to microprocessor module; Wireless communication module is used to receive the instruction and data that microprocessor module sends, and outwards reports information and orientation that human body is fallen.
Above-mentioned handles data analysis, judges whether human body falls, and is meant the following work of microprocessor module execution:
(1) make up based on the human body of the support vector machine behavior model of cognition of falling, its technical characterictic is identical with the method that above-mentioned a kind of monitoring is fallen, and repeats no more.
(2) receive described kinestate data in real time,, judge whether human body falls, and its technical characterictic is identical with the method that above-mentioned a kind of monitoring is fallen, and repeats no more by the behavior model of cognition of falling based on the human body of support vector machine.
Preferably, described sensor assembly is an acceleration transducer, described kinestate data are acceleration information, whether described judgement human body falls also comprises whether the trunk inclination angle being satisfied the judgement of inclination angle setting value and whether the persistent period that the inclination angle setting value is satisfied at the trunk inclination angle is reached the judgement of setting the persistent period, described trunk inclination angle is by obtaining described acceleration information computing.
Preferably, above-mentioned wireless communication module is a gsm module.
After the final decision human body is fallen, microprocessor module control sound alarm module sends alarm sound, in time sue and labour with the attention that causes peripheral people, microprocessor module also sends instruction and data to gsm module simultaneously, outwards report information and the orientation that human body is fallen by the control gsm module, warning message is sent to corresponding monitoring center and guardian.This function can be located the residing position of user fast, reduces the first aid difficulty greatly, and monitoring person and custodial person are had very big help.
Further, this monitoring device of falling can also comprise alarm units such as vibrations, flash of light.
If user is realized also comparatively clear-headed and had certain self-care ability in the back of falling, user can also manually be reported to the police by the input button on the device.
As shown in Figure 7, the device that a kind of monitoring of the present invention is fallen places waist in use, the reasons are as follows: 1. with respect to head, place waist to make things convenient for long periods of wear, less to the life interference, sense of discomfort is lower; 2. the hand motion complexity is difficult to extract useful information from hand motion changes; 3. in the past studies have shown that the center that pelvis can be considered as human body, so the variation of the accekeration of waist more can reflect gravity center of human body's variation.
Among the present invention, the type of falling that is adopted all is six kinds of types of typically falling, for other atypical the fall type or the influence of environment, adopt in the same way, gather the kinestate data of atypical type of falling, form the fall characteristic vector sample of type of atypia, after support vector machine is trained, obtain the fall standard feature vector of type of this atypia, add to advance in the template base getting final product.
The present invention can judge more exactly and whether fall, and the judgement of uniting by support vector machine pattern recognition and inclination angle threshold values has greatly improved and judges the accuracy of falling.
Below only statement for main principle of the present invention and spirit being carried out by preferred embodiment; be not limited to the present invention; for a person skilled in the art; the present invention can have various changes and variation; all any modifications of being done within the spirit and principles in the present invention, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (10)
1. the method that monitoring is fallen is characterized in that comprising the following steps:
1) makes up based on the human body of the support vector machine behavior model of cognition of falling;
2) the kinestate data of real-time human body by the behavior model of cognition of falling based on the human body of support vector machine, judge whether human body falls;
3) warning after judgement is fallen.
2. the method that a kind of monitoring according to claim 1 is fallen is characterized in that, described structure comprises step based on the human body of the support vector machine behavior model of cognition of falling:
1) construction feature sample: gather the kinestate data of the different behavioral pattern of human body in the setting-up time, carry out forming kinestate sample characteristics vector after the pretreatment construction feature vector storehouse;
2) training classifier: according to based on the support vector machine principle, be that the feature samples that the kinestate data of pattern constitute is trained with the human body different rows respectively, obtain being used for the fall supporting vector machine model of state of human body;
3) make up the model of cognition of falling: from the training characteristics sample, extract the standard feature vector that is used to discern the behavior of falling, make up the template base of falling.
3. the method for falling according to claim 1, one of 2 described a kind of monitorings is characterized in that described kinestate data are acceleration information.
4. the method that a kind of monitoring according to claim 3 is fallen is characterized in that, whether described judgement human body falls also comprises the judgement of whether the trunk inclination angle being satisfied the inclination angle setting value.
5. the method that a kind of monitoring according to claim 4 is fallen is characterized in that, whether described judgement human body falls comprises also whether the persistent period that inclination angle setting value is satisfied at the trunk inclination angle reaches the judgement of setting the persistent period.
6. the device that monitoring is fallen comprises microprocessor module and connected sensor assembly, audible alarm module, input button; Described sensor assembly is used for the kinestate data of human body; Described microprocessor module is used to receive described kinestate data, and data analysis is handled, and judges whether human body falls; Described audible alarm module is used to send alarm sound; Described input button is used for setting the mode of operation of microprocessor module to microprocessor module input external command and data; It is characterized in that described to the data analysis processing, whether the judgement human body is fallen and comprised:
1) makes up based on the human body of the support vector machine behavior model of cognition of falling;
2) receive described kinestate data in real time,, judge whether human body falls by the behavior model of cognition of falling based on the human body of support vector machine.
7. the device that a kind of monitoring according to claim 6 is fallen is characterized in that, described structure comprises based on the human body of the support vector machine behavior model of cognition of falling:
1) construction feature sample: gather the kinestate data of the different behavioral pattern of human body in the setting-up time, carry out forming kinestate sample characteristics vector after the pretreatment construction feature vector storehouse;
2) training classifier: according to based on the support vector machine principle, be that the feature samples that the kinestate data of pattern constitute is trained with the human body different rows respectively, obtain being used for the fall support vector machine detector of state of human body;
3) make up the model of cognition of falling: from the training characteristics sample, extract the standard feature vector that is used to discern the behavior of falling, make up the template base of falling.
8. the device of falling according to claim 6, one of 7 described a kind of monitorings, it is characterized in that, described sensor assembly is an acceleration transducer, described kinestate data are acceleration information, whether described judgement human body falls also comprises whether the trunk inclination angle being satisfied the judgement of inclination angle setting value and whether the persistent period that the inclination angle setting value is satisfied at the trunk inclination angle is reached the judgement of setting the persistent period, described trunk inclination angle is by obtaining described acceleration information computing.
9. the device of falling as claim 6, one of 7 described a kind of monitorings is characterized in that, also comprises the GPS module, the wireless communication module that are connected with described microprocessor module; Described GPS module is used to measure the residing position of human body, and sends position data to microprocessor module; Described wireless communication module is used to receive the instruction and data that microprocessor module sends, and outwards reports information and orientation that human body is fallen.
10. the device that a kind of monitoring as claimed in claim 9 is fallen is characterized in that, described wireless communication module is a gsm module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310174295XA CN103211599A (en) | 2013-05-13 | 2013-05-13 | Method and device for monitoring tumble |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310174295XA CN103211599A (en) | 2013-05-13 | 2013-05-13 | Method and device for monitoring tumble |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103211599A true CN103211599A (en) | 2013-07-24 |
Family
ID=48810016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310174295XA Pending CN103211599A (en) | 2013-05-13 | 2013-05-13 | Method and device for monitoring tumble |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103211599A (en) |
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103417219A (en) * | 2013-09-11 | 2013-12-04 | 重庆大学 | Wearable human body falling detection device |
| CN103606248A (en) * | 2013-09-30 | 2014-02-26 | 广州市香港科大霍英东研究院 | Automatic detection method and system for human body falling-over |
| CN103927851A (en) * | 2014-04-04 | 2014-07-16 | 余志峰 | Individualized multi-threshold tumble detection method and system |
| CN104586398A (en) * | 2013-10-30 | 2015-05-06 | 广州华久信息科技有限公司 | Old man falling detecting method and system based on multi-sensor fusion |
| CN104740720A (en) * | 2013-12-27 | 2015-07-01 | 亚星健康科技有限公司 | Medical container abnormality detecting method and apparatus |
| CN105310696A (en) * | 2015-11-06 | 2016-02-10 | 中国科学院计算技术研究所 | Fall detection model construction method as well as corresponding fall detection method and apparatus |
| CN105701965A (en) * | 2014-12-15 | 2016-06-22 | 朱保东 | Wearable safety equipment and method thereof |
| CN105760861A (en) * | 2016-03-29 | 2016-07-13 | 华东师范大学 | Epileptic seizure monitoring method and system based on depth data |
| CN105872962A (en) * | 2015-01-23 | 2016-08-17 | 上海爱戴科技有限公司 | Old people's tumble tracking system for old people |
| CN105930767A (en) * | 2016-04-06 | 2016-09-07 | 南京华捷艾米软件科技有限公司 | Human body skeleton-based action recognition method |
| CN106407732A (en) * | 2016-12-09 | 2017-02-15 | 珠海安润普科技有限公司 | Method, device and system for identifying states of target object and intelligent clothes |
| CN106407996A (en) * | 2016-06-30 | 2017-02-15 | 华南理工大学 | Machine learning based detection method and detection system for the fall of the old |
| CN106473265A (en) * | 2016-12-03 | 2017-03-08 | 石家庄学院 | A kind of intelligence is fallen and is guarded T-shirt and its fall detection algorithm |
| CN106781278A (en) * | 2017-02-04 | 2017-05-31 | 华南理工大学 | A kind of fall detection method and device based on Fusion |
| CN106778481A (en) * | 2016-11-15 | 2017-05-31 | 上海百芝龙网络科技有限公司 | A kind of body heath's monitoring method |
| CN106875630A (en) * | 2017-03-13 | 2017-06-20 | 中国科学院计算技术研究所 | A kind of wearable fall detection method and system based on hierarchical classification |
| CN106910314A (en) * | 2017-02-03 | 2017-06-30 | 同济大学 | A kind of personalized fall detection method based on the bodily form |
| CN106951852A (en) * | 2017-03-15 | 2017-07-14 | 深圳汇创联合自动化控制有限公司 | A kind of effective Human bodys' response system |
| CN106960541A (en) * | 2017-03-10 | 2017-07-18 | 浙江大学城市学院 | Adaptive multistage fall detection method |
| CN107358248A (en) * | 2017-06-07 | 2017-11-17 | 南京邮电大学 | A kind of method for improving fall detection system precision |
| CN107406243A (en) * | 2015-03-25 | 2017-11-28 | 意大利德纳有限责任公司 | For detecting the imminent system and method tumbled of vehicle |
| CN107644229A (en) * | 2017-08-31 | 2018-01-30 | 浙江工业大学 | Indoor anomaly detection method based on multi-sensor information fusion |
| CN108509897A (en) * | 2018-03-29 | 2018-09-07 | 同济大学 | A kind of human posture recognition method and system |
| CN108960056A (en) * | 2018-05-30 | 2018-12-07 | 西南交通大学 | A kind of fall detection method based on posture analysis and Support Vector data description |
| CN109009145A (en) * | 2018-07-24 | 2018-12-18 | 西安工程大学 | A kind of tumble judgment method based on wearable device |
| CN109119172A (en) * | 2018-07-12 | 2019-01-01 | 中原工学院 | A kind of human body behavioral value method based on ant colony algorithm |
| CN109886086A (en) * | 2019-01-04 | 2019-06-14 | 南京邮电大学 | Pedestrian detection method based on HOG feature and Linear SVM cascade classifier |
| CN110298276A (en) * | 2019-06-19 | 2019-10-01 | 北京工业大学 | Fall detection method based on convolutional neural networks |
| CN110390318A (en) * | 2019-07-30 | 2019-10-29 | 广东电网有限责任公司 | Falling down recognition methods, fall down identification insole and system applied to electric field |
| CN110443150A (en) * | 2019-07-10 | 2019-11-12 | 思百达物联网科技(北京)有限公司 | A kind of fall detection method, device, storage medium |
| CN110598536A (en) * | 2019-07-31 | 2019-12-20 | 杭州连帆科技有限公司 | Falling detection method and system based on human skeleton motion model |
| CN111436940A (en) * | 2020-03-20 | 2020-07-24 | 北京邮电大学 | Gait health assessment method and device |
| CN111887859A (en) * | 2020-08-05 | 2020-11-06 | 安徽华米智能科技有限公司 | Fall behavior recognition method and device, electronic device and medium |
| CN112235464A (en) * | 2019-06-28 | 2021-01-15 | 华为技术有限公司 | Falling detection-based help calling method and electronic equipment |
| CN112386249A (en) * | 2019-08-14 | 2021-02-23 | 中移(苏州)软件技术有限公司 | Fall detection method and device, equipment and storage medium |
| CN112466090A (en) * | 2020-11-26 | 2021-03-09 | 汤泽金 | Intelligent rescue bracelet |
| CN113856264A (en) * | 2020-06-30 | 2021-12-31 | 美卓奥图泰芬兰有限公司 | Pressure filter |
| WO2024036958A1 (en) * | 2022-08-17 | 2024-02-22 | 山东浪潮科学研究院有限公司 | Tinyml-based low-power-consumption intelligent assistance system |
| CN108652637B (en) * | 2018-06-30 | 2024-04-12 | 源珈力医疗器材国际贸易（上海）有限公司 | Wearable falling prediction protection system and prediction method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005237576A (en) * | 2004-02-25 | 2005-09-08 | Kansai Electric Power Co Inc:The | Tumble judgment device |
| CN101536053A (en) * | 2006-11-14 | 2009-09-16 | 皇家飞利浦电子股份有限公司 | System for fall prevention and a method for fall prevention using such a system |
| CN101615332A (en) * | 2008-06-24 | 2009-12-30 | 英华达股份有限公司 | Fall detection alarm device, method and use the shoes of this device |
| CN102610056A (en) * | 2012-03-16 | 2012-07-25 | 清华大学 | Mobile phone wearing mode-oriented falling event detection system and method |
-
2013
- 2013-05-13 CN CN201310174295XA patent/CN103211599A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005237576A (en) * | 2004-02-25 | 2005-09-08 | Kansai Electric Power Co Inc:The | Tumble judgment device |
| CN101536053A (en) * | 2006-11-14 | 2009-09-16 | 皇家飞利浦电子股份有限公司 | System for fall prevention and a method for fall prevention using such a system |
| CN101615332A (en) * | 2008-06-24 | 2009-12-30 | 英华达股份有限公司 | Fall detection alarm device, method and use the shoes of this device |
| CN102610056A (en) * | 2012-03-16 | 2012-07-25 | 清华大学 | Mobile phone wearing mode-oriented falling event detection system and method |
Cited By (52)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103417219A (en) * | 2013-09-11 | 2013-12-04 | 重庆大学 | Wearable human body falling detection device |
| CN103606248B (en) * | 2013-09-30 | 2016-08-10 | 广州市香港科大霍英东研究院 | A kind of falling over of human body automatic testing method and system |
| CN103606248A (en) * | 2013-09-30 | 2014-02-26 | 广州市香港科大霍英东研究院 | Automatic detection method and system for human body falling-over |
| CN104586398A (en) * | 2013-10-30 | 2015-05-06 | 广州华久信息科技有限公司 | Old man falling detecting method and system based on multi-sensor fusion |
| CN104740720A (en) * | 2013-12-27 | 2015-07-01 | 亚星健康科技有限公司 | Medical container abnormality detecting method and apparatus |
| CN104740720B (en) * | 2013-12-27 | 2017-09-12 | 亚星健康科技有限公司 | The sensing method for abnormal status and device of container for medical use |
| CN103927851B (en) * | 2014-04-04 | 2016-06-29 | 余志峰 | A kind of individualized multi thresholds fall detection method and system |
| CN103927851A (en) * | 2014-04-04 | 2014-07-16 | 余志峰 | Individualized multi-threshold tumble detection method and system |
| CN105701965A (en) * | 2014-12-15 | 2016-06-22 | 朱保东 | Wearable safety equipment and method thereof |
| CN105872962A (en) * | 2015-01-23 | 2016-08-17 | 上海爱戴科技有限公司 | Old people's tumble tracking system for old people |
| US10507826B2 (en) | 2015-03-25 | 2019-12-17 | Dana Italia Srl | System and method for detecting an impending tip over of a vehicle |
| CN107406243A (en) * | 2015-03-25 | 2017-11-28 | 意大利德纳有限责任公司 | For detecting the imminent system and method tumbled of vehicle |
| CN105310696A (en) * | 2015-11-06 | 2016-02-10 | 中国科学院计算技术研究所 | Fall detection model construction method as well as corresponding fall detection method and apparatus |
| CN105760861A (en) * | 2016-03-29 | 2016-07-13 | 华东师范大学 | Epileptic seizure monitoring method and system based on depth data |
| CN105930767A (en) * | 2016-04-06 | 2016-09-07 | 南京华捷艾米软件科技有限公司 | Human body skeleton-based action recognition method |
| CN105930767B (en) * | 2016-04-06 | 2019-05-17 | 南京华捷艾米软件科技有限公司 | A kind of action identification method based on human skeleton |
| CN106407996A (en) * | 2016-06-30 | 2017-02-15 | 华南理工大学 | Machine learning based detection method and detection system for the fall of the old |
| CN106778481A (en) * | 2016-11-15 | 2017-05-31 | 上海百芝龙网络科技有限公司 | A kind of body heath's monitoring method |
| CN106473265A (en) * | 2016-12-03 | 2017-03-08 | 石家庄学院 | A kind of intelligence is fallen and is guarded T-shirt and its fall detection algorithm |
| CN106407732A (en) * | 2016-12-09 | 2017-02-15 | 珠海安润普科技有限公司 | Method, device and system for identifying states of target object and intelligent clothes |
| CN106910314A (en) * | 2017-02-03 | 2017-06-30 | 同济大学 | A kind of personalized fall detection method based on the bodily form |
| CN106781278A (en) * | 2017-02-04 | 2017-05-31 | 华南理工大学 | A kind of fall detection method and device based on Fusion |
| CN106960541A (en) * | 2017-03-10 | 2017-07-18 | 浙江大学城市学院 | Adaptive multistage fall detection method |
| CN106875630A (en) * | 2017-03-13 | 2017-06-20 | 中国科学院计算技术研究所 | A kind of wearable fall detection method and system based on hierarchical classification |
| CN106875630B (en) * | 2017-03-13 | 2018-12-04 | 中国科学院计算技术研究所 | A kind of wearable fall detection method and system based on hierarchical classification |
| CN106951852A (en) * | 2017-03-15 | 2017-07-14 | 深圳汇创联合自动化控制有限公司 | A kind of effective Human bodys' response system |
| CN107358248A (en) * | 2017-06-07 | 2017-11-17 | 南京邮电大学 | A kind of method for improving fall detection system precision |
| CN107358248B (en) * | 2017-06-07 | 2021-03-23 | 南京邮电大学 | Method for improving falling detection system precision |
| CN107644229A (en) * | 2017-08-31 | 2018-01-30 | 浙江工业大学 | Indoor anomaly detection method based on multi-sensor information fusion |
| CN108509897A (en) * | 2018-03-29 | 2018-09-07 | 同济大学 | A kind of human posture recognition method and system |
| CN108960056A (en) * | 2018-05-30 | 2018-12-07 | 西南交通大学 | A kind of fall detection method based on posture analysis and Support Vector data description |
| CN108652637B (en) * | 2018-06-30 | 2024-04-12 | 源珈力医疗器材国际贸易（上海）有限公司 | Wearable falling prediction protection system and prediction method thereof |
| CN109119172A (en) * | 2018-07-12 | 2019-01-01 | 中原工学院 | A kind of human body behavioral value method based on ant colony algorithm |
| CN109119172B (en) * | 2018-07-12 | 2020-06-05 | 中原工学院 | Human behavior detection method based on bee colony algorithm |
| CN109009145A (en) * | 2018-07-24 | 2018-12-18 | 西安工程大学 | A kind of tumble judgment method based on wearable device |
| CN109886086B (en) * | 2019-01-04 | 2020-12-04 | 南京邮电大学 | Pedestrian detection method based on HOG (histogram of oriented gradient) features and linear SVM (support vector machine) cascade classifier |
| CN109886086A (en) * | 2019-01-04 | 2019-06-14 | 南京邮电大学 | Pedestrian detection method based on HOG feature and Linear SVM cascade classifier |
| CN110298276A (en) * | 2019-06-19 | 2019-10-01 | 北京工业大学 | Fall detection method based on convolutional neural networks |
| US11928947B2 (en) | 2019-06-28 | 2024-03-12 | Huawei Technologies Co., Ltd. | Fall detection-based help-seeking method and electronic device |
| CN112235464B (en) * | 2019-06-28 | 2022-05-31 | 华为技术有限公司 | Falling detection-based help calling method and electronic equipment |
| CN112235464A (en) * | 2019-06-28 | 2021-01-15 | 华为技术有限公司 | Falling detection-based help calling method and electronic equipment |
| CN110443150A (en) * | 2019-07-10 | 2019-11-12 | 思百达物联网科技(北京)有限公司 | A kind of fall detection method, device, storage medium |
| CN110390318A (en) * | 2019-07-30 | 2019-10-29 | 广东电网有限责任公司 | Falling down recognition methods, fall down identification insole and system applied to electric field |
| CN110598536A (en) * | 2019-07-31 | 2019-12-20 | 杭州连帆科技有限公司 | Falling detection method and system based on human skeleton motion model |
| CN112386249A (en) * | 2019-08-14 | 2021-02-23 | 中移(苏州)软件技术有限公司 | Fall detection method and device, equipment and storage medium |
| CN112386249B (en) * | 2019-08-14 | 2023-09-19 | 中移(苏州)软件技术有限公司 | Fall detection method and device, equipment and storage medium |
| CN111436940A (en) * | 2020-03-20 | 2020-07-24 | 北京邮电大学 | Gait health assessment method and device |
| CN113856264A (en) * | 2020-06-30 | 2021-12-31 | 美卓奥图泰芬兰有限公司 | Pressure filter |
| CN113856264B (en) * | 2020-06-30 | 2024-05-14 | 美卓奥图泰芬兰有限公司 | Pressure filter |
| CN111887859A (en) * | 2020-08-05 | 2020-11-06 | 安徽华米智能科技有限公司 | Fall behavior recognition method and device, electronic device and medium |
| CN112466090A (en) * | 2020-11-26 | 2021-03-09 | 汤泽金 | Intelligent rescue bracelet |
| WO2024036958A1 (en) * | 2022-08-17 | 2024-02-22 | 山东浪潮科学研究院有限公司 | Tinyml-based low-power-consumption intelligent assistance system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103211599A (en) | Method and device for monitoring tumble | |
| Hsu et al. | Human daily and sport activity recognition using a wearable inertial sensor network | |
| CN105125221B (en) | Detecting system and method are fallen down in cloud service in real time | |
| He et al. | A low power fall sensing technology based on FD-CNN | |
| CN105310696B (en) | A kind of fall detection model building method and corresponding fall detection method and device | |
| CN104361321B (en) | A kind of method for judging the elderly and falling down behavior and balance ability | |
| CN111932828A (en) | Intelligent old-age care monitoring and early warning system based on digital twin technology | |
| CN105118236A (en) | Paralysis falling detection and prevention device and processing method thereof | |
| CN104269025B (en) | Wearable single node feature and the position choosing method of monitoring is fallen down towards open air | |
| CN205050303U (en) | Human paralysis of intelligence is monitoring devices | |
| CN106725445B (en) | A kind of the portable body injury gained in sports monitor system and method for brain wave control | |
| CN109009145A (en) | A kind of tumble judgment method based on wearable device | |
| CN106407996A (en) | Machine learning based detection method and detection system for the fall of the old | |
| CN111325946A (en) | Fall detection method and system based on edge calculation | |
| Malshika Welhenge et al. | Human activity classification using long short-term memory network | |
| Yang et al. | Fall detection system based on infrared array sensor and multi-dimensional feature fusion | |
| CN105869354B (en) | A kind of Falls Among Old People detection method based on attractor propagation algorithm | |
| Ding et al. | Energy efficient human activity recognition using wearable sensors | |
| Medrano et al. | Personalizable smartphone application for detecting falls | |
| CN107146378A (en) | A kind of human body tumble decision method and device | |
| CN107019501B (en) | Remote tumble detection method and system based on genetic algorithm and probabilistic neural network | |
| Shi et al. | Fall detection system based on inertial mems sensors: Analysis design and realization | |
| KR102337861B1 (en) | Apparatus for fall prediction using lead time and method thereof | |
| Wang et al. | Comparison of four machine learning algorithms for a pre-impact fall detection system | |
| CN103632133A (en) | Human gesture recognition method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C05 | Deemed withdrawal (patent law before 1993) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130724 |