CN105469546A - Tumbling alarm system and method - Google Patents

Abstract

The invention relates to a tumbling alarm system and method. The system comprises a single-chip microcomputer which is internally provided with a programming module capable of determining a tumbling state; the data input terminal of the single-chip microcomputer is in connection with an inertia data acquisition unit and a positioning receiving unit; the data output terminal of the single-chip microcomputer is in connection with an audio frequency unit and a wireless communication unit; the wireless communication unit is in connection with an assigned terminal, and sends a tumbling state of a user to be detected and position information to the assigned terminal. The method comprises the steps of: collecting the acceleration vectors and angular velocity vectors of the user to be detected; filtering collected acceleration vectors and angular velocity vectors; synthesizing sampling acceleration vectors, performing threshold determination on a synthesized acceleration, and preliminarily determining a possible tumbling state; performing sensitive treatment on the sampling acceleration vectors and sampling angular velocity vectors; setting a sampling window; performing fuzzification treatment on the sampling value in the sampling window to determine tumbling or not; and transmitting a tumbling signal to the assigned terminal to realize tumbling alarm.

Description

A kind of tumbling alarm system and method

Technical field

The present invention relates to a kind of tumbling alarm system and method, particularly relate to a kind of tumbling alarm system and method that utilize fuzzy self-adaption to detect to fall down.

Background technology

Former generation is along with the increase at age, and health and mental function are progressively degenerated, and the incidence of Falls in Old People is high, consequence serious, if old man falls down rear generation stupor, so situation is just more critical.This also becomes the injury of aged first place or the cause of death.Fall in population of China to kill in because of cis-position in accidental wound and come the 4th, in the elderly of over-65s, then occupy first place, and sharply rise with the mortality ratio that the increase at age is fallen, in the elderly of more than 85 years old, reach peak.Aly old man's real-time status can be accurately detected and timely state of old man being fallen down informs that the device of household is very necessary so design.

In the prior art, fall down detection technique for inertia and generally have two kinds of modes, a kind of for adopting the method for neural network filter, these class methods need more hardware resource, higher to the requirement of hardware process speed, and set up neural network for different users, process is complicated, and reproducibility is poor; Another kind is by 3-axis acceleration vector delivery, and sets the method that threshold decision falls down, and these class methods are simple to operation, real-time, but judges too simple, and False Rate is high.

Summary of the invention

In order to overcome the above problems, the invention provides a kind of tumbling alarm system and method, this system and method is simple to operation, real-time, and judges that precision is high.

To achieve these goals, the present invention adopts following technical scheme:

A kind of tumbling alarm system, comprises single-chip microcomputer, stores the programming module that can determine whether the state of falling down in single-chip microcomputer, and the data input pin of described single-chip microcomputer connects inertial data collecting unit, and position receiver unit; The data output end of described single-chip microcomputer is connected with audio unit and radio communication unit, and radio communication unit is connected with designated terminal, by person to be measured fall down state and positional information is sent to designated terminal.

Described inertial data collecting unit comprises the acceleration transducer gathering acceleration signal, and the gyroscope of acquisition angle rate signal.

Low-pass filter is provided with, the inertial data filtering that low-pass filter will gather in inertial data collecting unit, with exclusive PCR data between inertial data collecting unit and single-chip microcomputer.

Tumbling alarm system also comprises the display button unit that this warning system of operation is run, this unit comprises display screen, the below of display screen is provided with multiple setting button, also be provided with alarm key in emergency situations between display screen and setting button, the both sides of alarm key are respectively arranged with confirming button and negative button.

One falls down alarm method, comprises the following steps:

1) gather the acceleration on person to be measured three directions and angular velocity, and form vector acceleration and angular velocity vector;

2) to the vector acceleration collected and angular velocity vector filtering, sampled acceleration vector and sampling angular velocity vector is formed;

3) synthesis of sampled acceleration vector is formed resultant acceleration, and carry out threshold decision to resultant acceleration, whether preliminary judgement is possible fall down state; Resultant acceleration is compared with setting threshold value, if resultant acceleration is not less than setting threshold value, then judges that person to be measured is normal; If resultant acceleration is less than setting threshold value, then judge that person to be measured may fall down;

4) by sampled acceleration vector and sampling angular velocity vector sensitization process;

5) sample window synthesis sampling matrix is set;

6) to the sampling matrix Fuzzy processing in sample window to determine whether to fall down;

7) Signal transmissions will be fallen down to designated terminal to realize falling down warning.

Described step 3) in the setting concrete grammar of threshold value comprise the following steps:

A. at the multiple time section of every Japan-China division;

B. the amplitude of the resultant acceleration in each time section is recorded;

C. in accumulative each time section, resultant acceleration amplitude exceedes the quantity of preset value;

D. the section movement intensity in each time section is determined according to accumulated quantity;

E. determine to set threshold value according to following formula:

T _acc=T _exp+ a (E-E _exp), wherein T _accrepresent setting threshold value, unit is m/s ²; T _expthe threshold preset parameter and default magnification ratio coefficient that obtain through overtesting is respectively with parameter a.E is section movement intensity, E _expfor the parameter preset of section movement intensity.

Described step 4) be specially: by sampled acceleration vector and sampling angular velocity vector for the low Data Synthesis reprocessing of the fall events degree of association or reject; The data high for the degree of association in fall events amplify process.

If judged result is for falling down in threshold determination, will judge by toggle window, window judges to receive the sampling matrix from matrixing processing unit.

Described step 5) in set sample window to synthesize sampling matrix detailed process as follows: threshold determination is as after falling down, from the sampling time point obtaining next sampled acceleration and sampling angle number vector, a window is set up to judge sampling time point t at regular intervals _j, set up m altogether, t ₁t ₂Λ t _m, judge that sampling time point gathers the n dimension data from data sensitive processing unit at each window, form sampling matrix S:

$S=\left[\begin{array}{cccc}{S}_{11}& S_{12}& \mathrm{\Λ}& S_{1m}\\ S_{21}& S_{22}& \mathrm{\Λ}& S_{2m}\\ M& & & M\\ S_{n1}& S_{n2}& \mathrm{\Λ}& S_{nm}\end{array}\right].$

Wherein S _ijrepresent the element in sampling matrix S.The element representation of each row in matrix is at a corresponding time point t _jn dimension data (the S from data sensitive processing unit that place collects _1js _2jΛ S _nj) ^t.

Described step 6) to the sampling matrix Fuzzy processing in sample window to determine whether to fall down, specifically comprise the following steps:

Utilize following formula to the sample magnitude Fuzzy processing in sample window:

$F=\left\{\begin{array}{cc}{N}_{w1}& 0\&le;|S_{ij}-S_{ij}^{\exp }|\&le;T_{w1}\\ N_{w2}& T_{w1}<|S_{ij}-S_{ij}^{\exp }|\&le;T_{w2}\\ N_{w3}& T_{w2}<|S_{ij}-S_{ij}^{\exp }|\&le;T_{w3}\\ 0& T_{w3}<|S_{ij}-S_{ij}^{\exp }|\end{array}\right.,$

Wherein be through each some S in the sampling matrix S testing and determine _ijexpectation value, so have the different expectation value of m × n $\left[\begin{array}{cccc}{S}_{11}^{\exp }& S_{12}^{\exp }& \mathrm{\&Lambda;}& S_{1m}^{\exp }\\ S_{21}^{\exp }& S_{22}^{\exp }& \mathrm{\&Lambda;}& S_{2m}^{\exp }\\ M& & & M\\ S_{n1}^{\exp }& S_{n2}^{\exp }& \mathrm{\&Lambda;}& S_{nm}^{\exp }\end{array}\right];$ N _w1, N _w2and N _w3for the fuzzy value in fuzzification function F.T _w1, T _w2and T _w3for the parameter preset in fuzzification function F.Point S in sampling matrix _ijthe result obtained after input function F is N _ij.M × n in the fuzzy matrix formed after above-mentioned Fuzzy processing numerical value is added synthesis judgement and falls down Parameter N, judgement is fallen down Parameter N and fall down Parameter N with default _winrelatively, as N > N _wintime, be judged to fall down, as N < N _winin time, is judged to not fall down.

Compared with prior art, the present invention has following outstanding substantive distinguishing features and significant advantage:

(1) method of operating of present system is simple, real-time;

(2) the inventive method is by the method that adaptive threshold judges and self-adapting window judges, and regulates adding intensive treatment and arranging obfuscation self application of responsive vector, and make judgement precision of the present invention higher, False Rate is less.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of tumbling alarm system in the present invention.

Fig. 2 is the display interface schematic diagram of tumbling alarm system in the present invention.

Fig. 3 is the process flow diagram falling down alarm method in the present invention.

Fig. 4 is the change schematic diagram of amplitude in a time section in the present invention.

Embodiment

Below in conjunction with accompanying drawing, specific embodiments of the invention are specifically described.

As shown in Figure 1, a kind of tumbling alarm system, comprises single-chip microcomputer, stores the programming module that can determine whether the state of falling down in single-chip microcomputer, and the data input pin of described single-chip microcomputer connects inertial data collecting unit, and position receiver unit; The data output end of described single-chip microcomputer is connected with audio unit and radio communication unit, and radio communication unit is connected with designated terminal, by person to be measured fall down state and positional information is sent to designated terminal.

Above-mentioned inertial data collecting unit comprises the acceleration transducer that can gather acceleration signal, and can the gyroscope of acquisition angle rate signal.

Be provided with low-pass filter between inertial data collecting unit and single-chip microcomputer, the inertial data filtering that low-pass filter can will gather in inertial data collecting unit, the noise signal of high frequency filters by low-pass filter, eliminates interfering data.

In addition, as shown in Figure 1, this tumbling alarm system also includes the audio unit be connected with single-chip microcomputer signal output part, and audio unit can send alerting signal, when person to be measured falls down, by the warning of sound, causes the concern of people around, thus asks for help; Also can in audio unit typing sound, arrive when taking medicine the time, remind user to take medicine the quantity etc. of kind.

As shown in Figure 2, in the present invention, tumbling alarm system also comprises the display button unit that can operate this warning system and run, this unit comprises and can show date, time, take medicine arrange display screen 1, the below of display screen 1 is provided with multiple setting button 2, the numerical value preset in above-mentioned single-chip microcomputer specifically can be changed by different buttons, simultaneously, alarm key 5 in emergency situations is also provided with between display screen 1 and setting button 2, the both sides of alarm key 5 are respectively arranged with confirming button 3 and negative button 4, when specifically judging, as system send fall down alarm time, person to be measured is non-state of falling down, then can press negative button 4, by negative button 4, error warning can be carried out to default value in this tumbling alarm system, after error is reported to the police, the program that can start in single-chip microcomputer restarts setting, improve the accuracy of falling down judgement.

As shown in Figure 3, alarm method of falling down of the present invention comprises the following steps:

1) gather the acceleration on person to be measured three directions and angular velocity, and form vector acceleration V _a(X, Y, Z) and angular velocity vector V _g(X, Y, Z);

2) sampled acceleration vector V is formed to the vector acceleration collected and angular velocity vector filtering _acc(X, Y, Z) and sampling angular velocity vector V _gyr(X, Y, Z);

3) synthesis of sampled acceleration vector is formed resultant acceleration Δ _acc, and threshold decision is carried out to resultant acceleration, whether preliminary judgement may be the state of falling down;

4) by sampled acceleration vector and sampling angular velocity vector sensitization process;

5) sample window synthesis sampling matrix is set;

6) to the sample magnitude Fuzzy processing in sample window, parameter is fallen down in synthesis judgement, and falls down parameter and compare to determine whether to fall down with presetting;

7) Signal transmissions will be fallen down to designated terminal to realize falling down warning.

For said method, being described in detail as follows each step:

According to the position setting X, Y, Z axis coordinate of person to be measured, then the acceleration recorded and angular velocity vector are V _a(X, Y, Z) and V _g(X, Y, Z);

Shown in Fig. 3, sampled acceleration vector and sampling angular velocity vector are formed to the vector acceleration collected and angular velocity vector filtering, V _acc(X, Y, Z) and V _gyr(X, Y, Z).

Continue see Fig. 3, the synthesis of sampled acceleration vector is formed resultant acceleration, and threshold decision is carried out to resultant acceleration, preliminary judgement whether may for falling down state time, be specially, resultant acceleration is compared with setting threshold value, if resultant acceleration is not less than setting threshold value, then judges that person to be measured is normal; If resultant acceleration is less than setting threshold value, then judge that person to be measured may fall down.

The concrete grammar of above-mentioned middle setting threshold value comprises the following steps: at the multiple time section of every Japan-China division; Record the amplitude in each time section; In accumulative each time section, amplitude exceedes the quantity of preset value; The exercise intensity in each time section is determined according to accumulated quantity; Determine to set threshold value: T according to following formula _acc=T _exp+ a (E-E _exp), wherein T _accrepresent setting threshold value, unit is m/s ²; T _expthe threshold preset parameter and default magnification ratio coefficient that obtain through overtesting is respectively with parameter a.E is section movement intensity, E _expfor the parameter preset of section movement intensity.As shown in Figure 3, section movement Strength co-mputation unit collection acceleration information carries out calculating threshold value T in rear Automatic adjusument threshold decision _accsize.

Then be described as follows further for above-mentioned threshold decision step:

To extract in vector of samples acceleration three number of axle according to according to following formula synthesize, to resultant acceleration Δ _acccarry out threshold decision, work as Δ _acc<T _acc(wherein T _accfor setting threshold value) time, be judged as not falling down, then these data are given up automatically; Work as Δ _acc>T _accin time, is judged as falling down, and by the judgement of threshold value, achieves and preliminary falls down judgement.It should be noted that, because of threshold value T _accthere is different values time periods different in one day, and the size of these values is (as Fig. 3) of determining according to the strength information E of following acceleration and angular velocity.Lift a specific embodiment to be described, in one day, be divided into 12 time periods, be namely divided into 6:00-8:00,8:00-10:00,10:00-12:00 equal time section, be i.e. corresponding 12 threshold value T _acc1Λ T _acc12.The setting of these threshold values can carry out adaptive adjustment according to the service condition between user's operating period.

Specifically see Fig. 4, the time of one day 24 hours is divided into some time section Q ₁Λ Q _n, the strength information recording user's acceleration in each time section is E ₁Λ E _n, strength information E _iacquisition be at corresponding time section Q _iin time cumulation is carried out to the signal that amplitude exceedes certain value, by the length of time cumulation, be set in the intensity of motion in this time period, formula is: E _i=T _a1+ T _a2+ Λ+T _an.

Illustrate, in the time section of 6:00 to 8:00, i.e. Q in Fig. 4 _iin, strength information is E _i.The time of acceleration amplitude between A1 and A2 is T _a1, the time that acceleration amplitude is greater than A2 is T _a2.Exercise intensity E so in whole time section _i=T _a1+ T _a2.By the above-mentioned above-mentioned threshold value T of exercise intensity self-adaptative adjustment determined _accsize (as in Fig. 3), i.e. T _acc=T _exp+ a (E-E _exp).Wherein T _expfor threshold preset parameter, E _expfor section movement intensity parameter preset, a is for presetting magnification ratio coefficient.

Threshold decision ensures that all situations of falling down all can by judging, and carries out preliminary screening, to improve the work efficiency of detection for fuzzy judgement afterwards.Therefore, threshold determination is that ground floor judges, but the accuracy rate that this layer judges is lower, as long as because the possibility of falling down a little a little will be judged by this layer, based on this, the follow-up judgement of the present invention's setting is more accurate, not only increase work efficiency like this, also improve the accuracy of judgement.

Continue see Fig. 3, during by sampled acceleration vector and sampling angular velocity vector sensitization process, be specially Data Synthesis reprocessing low for the fall events degree of association in sampled acceleration is vectorial and sampling angular velocity vector or reject; Amplification process is carried out for the data that the degree of association in fall events is high.

Data selection for fall events degree of association height described herein is according to being: it is high that the angular velocity that person to be measured produces in the process of falling down and acceleration change data different in nature larger with the difference in change under normal condition are the degree of association, otherwise, be then that the degree of association is low.Illustrate, people in normal state, it is comparatively large that the acceleration (i.e. z-axis acceleration) of vertical direction changes fluctuation ratio, and when falling down, the acceleration fluctuation of vertical direction is larger equally, so these data of acceleration of vertical direction and the degree of association of fall events just smaller.On the contrary, in normal state, the acceleration change in xy face is smaller for people, and the acceleration change of falling down under state in xy face is larger, obviously can distinguish, then think that the data of x-axis and y-axis are high with the degree of association of fall events.

In addition, the Data Synthesis reprocessing low for the fall events degree of association or reject, lift a specific embodiment to be described as follows: as implied above, the acceleration in z-axis direction, normal fluctuation range under normal condition is-2g to 2g, the fluctuation range of falling down under state is-3g to 3g, under two states, its maximum difference is 1g, namely z-axis direction is shown low for the fall events degree of association, this kind of mode so can be adopted to arrange data, acceleration by z-axis is multiplied by bigger numerical, and (this numerical value can be selected arbitrarily, be excellent with 8-10), one that so just can reach larger numerical value change scope, thus increase the may differentiate of two states.

Data after sensitization process become n dimension from 6 dimensions.

If judged result is for falling down in threshold determination, will judge by toggle window.Window judges the sampling matrix that first can receive from matrixing processing unit.The detailed process that sampling matrix obtains is: threshold determination is for after falling down, and start from the sampling time point obtaining next sampled acceleration and sampling angle number vector, the sampling time puts t to set up a window to judge at regular intervals _j, set up m altogether, so the sampling time that window judges puts as t ₁t ₂Λ t _m, judge that sampling time point gathers the n dimension data from data sensitive processing unit at each window, form sampling matrix S:

$S=\left[\begin{array}{cccc}{S}_{11}& S_{12}& \mathrm{\&Lambda;}& S_{1m}\\ S_{21}& S_{22}& \mathrm{\&Lambda;}& S_{2m}\\ M& & & M\\ S_{n1}& S_{n2}& \mathrm{\&Lambda;}& S_{nm}\end{array}\right],$

Wherein S _ijrepresent the element in sampling matrix S.The element representation of each row in matrix is at a corresponding time point t _jn dimension data (the S from data sensitive processing unit that place collects _1js _2jΛ S _nj) ^t, m window judges that the time interval of point in sampling time is different.Window judges that the determination of position of sampling time point is according to the variation characteristic of human body under the state of falling down as basis.Because people's Changing Pattern of different time points in the process of falling down is different, then the window of corresponding selection judges that sampling time point is also different, inertial data (acceleration and angular velocity) change as human body in 0.1 second after falling down is violent, then the window of setting in this period judges that sampling time point should more crypto set, within the time period of 0.3-0.9 second of falling over of human body, the inertial data variation range of human body is less, and window judges that sampling time point can be more sparse.If window judges sampling time point m altogether, the n dimension data from data sensitive processing unit that each time point collects, namely forms the sampling matrix S that above-mentioned m × n ties up.

To the sampling matrix S Fuzzy processing in sample window.Specifically comprise the following steps: utilize following formula to sampling matrix S Fuzzy processing:

$F=\left\{\begin{array}{cc}{N}_{w1}& 0\&le;|S_{ij}-S_{ij}^{\exp }|\&le;T_{w1}\\ N_{w2}& T_{w1}<|S_{ij}-S_{ij}^{\exp }|\&le;T_{w2}\\ N_{w3}& T_{w2}<|S_{ij}-S_{ij}^{\exp }|\&le;T_{w3}\\ 0& T_{w3}<|S_{ij}-S_{ij}^{\exp }|\end{array}\right.,$

Wherein be through each some S in the sampling matrix S testing and determine _ijexpectation value, so have the different expectation value of m × n $\left[\begin{array}{cccc}{S}_{11}^{\exp }& S_{12}^{\exp }& \mathrm{\&Lambda;}& S_{1m}^{\exp }\\ S_{21}^{\exp }& S_{22}^{\exp }& \mathrm{\&Lambda;}& S_{2m}^{\exp }\\ M& & & M\\ S_{n1}^{\exp }& S_{n2}^{\exp }& \mathrm{\&Lambda;}& S_{nm}^{\exp }\end{array}\right];$ N _w1, N _w2and N _w3for the fuzzy value in fuzzification function F.T _w1, T _w2and T _w3for the parameter preset in fuzzification function F, in order to judge span.Point S in sampling matrix _ijthe result obtained after input function F is N _ij, N _ijvalue have four kinds may, be respectively N _w1, N _w2, N _w3or 0.Obfuscation matrix is obtained by after sampling matrix S input function F:

$\left[\begin{array}{cccc}{N}_{11}& N_{12}& \mathrm{\&Lambda;}& N_{1m}\\ N_{21}& N_{22}& \mathrm{\&Lambda;}& N_{2m}\\ M& & & NM\\ N_{n1}& N_{n2}& \mathrm{\&Lambda;}& {\mathrm{\&Lambda;}}_{nm}\end{array}\right]$

M × n in the fuzzy matrix formed after above-mentioned Fuzzy processing numerical value is added synthesis judgement and falls down Parameter N, N=N ₁₁+ Λ N _ij+ Λ+N _nm(i=1,2, Λ, nj=1,2, Λ, m).

Judgement is fallen down parameter and fall down Parameter N with default _winrelatively, fall down parameter as judged and be greater than default judgement when falling down parameter, be judged to be the state of falling down; As judge fall down parameter be less than preset fall down parameter time, be judged to be the state of not falling down, namely as N > N _wintime, be judged to fall down, as N < N _winin time, is judged to not fall down.

T in above-mentioned function F _w1, T _w2and T _w3for variable, its value according to fall down erroneous judgement information self-adapting adjustment, namely when system be mistaken for fall down attitude time, user presses erroneous judgement button, and system is judged as erroneous judgement state.Adaptation module will regulate T _w1, T _w2and T _w3size, accumulative namely along with erroneous judgement number of times, T _w1, T _w2and T _w3value more will meet the exercise habit of special user, make judgement stricter.

Claims (9)

1. a tumbling alarm system, is characterized in that, comprises single-chip microcomputer, stores the programming module that can determine whether the state of falling down in single-chip microcomputer, and the data input pin of described single-chip microcomputer connects inertial data collecting unit, and position receiver unit; The data output end of described single-chip microcomputer is connected with audio unit and radio communication unit, and radio communication unit is connected with designated terminal, by person to be measured fall down state and positional information is sent to designated terminal.

2. tumbling alarm system according to claim 1, is characterized in that, described inertial data collecting unit comprises the acceleration transducer gathering acceleration signal, and the gyroscope of acquisition angle rate signal.

3. tumbling alarm system according to claim 1, is characterized in that, be provided with low-pass filter between inertial data collecting unit and single-chip microcomputer, the inertial data filtering that low-pass filter will gather in inertial data collecting unit, with exclusive PCR data.

4. tumbling alarm system according to claim 1, it is characterized in that, tumbling alarm system also comprises the display button unit that this warning system of operation is run, this unit comprises display screen (1), the below of display screen (1) is provided with multiple setting button (2), also be provided with alarm key (5) in emergency situations between display screen (1) and setting button (2), the both sides of alarm key (5) are respectively arranged with confirming button (3) and negative button (4).

5. fall down an alarm method, it is characterized in that, comprise the following steps:

1) gather the acceleration on person to be measured three directions and angular velocity, and form vector acceleration and angular velocity vector;

2) to the vector acceleration collected and angular velocity vector filtering, sampled acceleration vector and sampling angular velocity vector is formed;

3) synthesis of sampled acceleration vector is formed resultant acceleration, and carry out threshold decision to resultant acceleration, whether preliminary judgement is possible fall down state; Resultant acceleration is compared with setting threshold value, if resultant acceleration is not less than setting threshold value, then judges that person to be measured is normal; If resultant acceleration is less than setting threshold value, then judge that person to be measured may fall down;

4) by sampled acceleration vector and sampling angular velocity vector sensitization process;

5) sample window synthesis sampling matrix is set;

6) to the sampling matrix Fuzzy processing in sample window to determine whether to fall down;

7) Signal transmissions will be fallen down to designated terminal to realize falling down warning.

6. fall down alarm method according to claim 5, it is characterized in that, described step 3) in the setting concrete grammar of threshold value comprise the following steps:

A. at the multiple time section of every Japan-China division;

B. the amplitude of the resultant acceleration in each time section is recorded;

C. in accumulative each time section, resultant acceleration amplitude exceedes the quantity of preset value;

D. the section movement intensity in each time section is determined according to accumulated quantity;

E. determine to set threshold value according to following formula:

T _acc=T _exp+ a (E-E _exp), wherein T _accrepresent setting threshold value, unit is m/s ²; T _expbe respectively with parameter a the threshold preset parameter and default magnification ratio coefficient that obtain through overtesting, E is section movement intensity, E _expfor the parameter preset of section movement intensity.

7. fall down alarm method according to claim 5, it is characterized in that, described step 4) be specially: by sampled acceleration vector and sampling angular velocity vector for the low Data Synthesis reprocessing of the fall events degree of association or reject; The data high for the degree of association in fall events amplify process.

8. fall down alarm method according to claim 5, it is characterized in that, if judged result is for falling down in threshold determination, will judge by toggle window, window judges to receive the sampling matrix from matrixing processing unit; Described step 5) in set sample window to synthesize sampling matrix detailed process as follows: threshold determination is as after falling down, from the sampling time point obtaining next sampled acceleration and sampling angle number vector, a window is set up to judge sampling time point t at regular intervals _j, set up m altogether, t ₁t ₂Λ t _m, judge that sampling time point gathers the n dimension data from data sensitive processing unit at each window, form sampling matrix S:

$S=\left[\begin{array}{cccc}{S}_{11}& S_{12}& \mathrm{\&Lambda;}& S_{1m}\\ S_{21}& S_{22}& \mathrm{\&Lambda;}& S_{2m}\\ M& & & M\\ S_{n1}& S_{n2}& \mathrm{\&Lambda;}& S_{nm}\end{array}\right]$

Wherein S _ijrepresent the element in sampling matrix S, the element representation of each row in matrix is at a corresponding time point t _jn dimension data (the S from data sensitive processing unit that place collects _1js _2jΛ S _nj) ^t.

9. fall down alarm method according to claim 5, it is characterized in that, described step 6) to the sampling matrix Fuzzy processing in sample window to determine whether to fall down, specifically comprise the following steps:

Utilize following formula to the sample magnitude Fuzzy processing in sample window:

$F=\left\{\begin{array}{cc}{N}_{w1}& 0\&le;\left|S_{ij}-S_{ij}^{\exp }\right|\&le;T_{w1}\\ N_{w2}& T_{w1}<\left|S_{ij}-S_{ij}^{\exp }\right|\&le;T_{w2}\\ N_{w3}& T_{w2}<\left|S_{ij}-S_{ij}^{\exp }\right|\&le;T_{w3}\\ 0& T_{w3}<\left|S_{ij}-S_{ij}^{\exp }\right|\end{array}\right.,$

Wherein be through each some S in the sampling matrix S testing and determine _ijexpectation value, so have the different expectation value of m × n $\left[\begin{array}{cccc}{S}_{11}^{\exp }& S_{12}^{\exp }& \mathrm{\&Lambda;}& S_{1m}^{\exp }\\ S_{21}^{\exp }& S_{22}^{\exp }& \mathrm{\&Lambda;}& S_{2m}^{\exp }\\ M& & & M\\ S_{n1}^{\exp }& S_{n2}^{\exp }& \mathrm{\&Lambda;}& S_{\mathrm{nm}}^{\exp }\end{array}\right];$ N _w1, N _w2and N _w3for the fuzzy value in fuzzification function F, T _w1, T _w2and T _w3for the parameter preset in fuzzification function F, the some S in sampling matrix _ijthe result obtained after input function F is N _ij, the m × n in the fuzzy matrix formed numerical value is added synthesis judgement and falls down Parameter N, judgement is fallen down Parameter N and fall down Parameter N with default after above-mentioned Fuzzy processing _winrelatively, as N > N _wintime, be judged to fall down, as N < N _winin time, is judged to not fall down.