CN107146377A - Pre-collision fall detection method and device - Google Patents

Abstract

The invention discloses a fall detection method and device before a collision. The method includes: judging whether the fall detection index time series is autocorrelated; if not, using the fall detection index time series of the detected individual to establish a statistical process control model; Correlation, process the fall detection index time series through the ARIMA model, and output non-autocorrelation data; establish a statistical process control model based on the non-autocorrelation data processed by the ARIMA model; judge whether the human body has fallen according to the statistical process control model, and if it is determined to fall, A fall alarm is issued. This program first establishes a statistical process control chart model, and uses this model to judge whether human activities are falls, taking into account the differences in the actual falls of different users, which can improve the accuracy of fall detection before collision; and can quickly detect the risk of falls occur.

Description

Pre-collision fall detection method and device

technical field

The invention relates to the field of fall state detection, in particular to a fall detection method and device before collision.

Background technique

Falls have always been an important problem threatening the health of the elderly. According to the report of the World Health Organization (WHO), nearly one-third of the elderly over the age of 65 experience at least one fall every year. In my country, falls are the primary cause of injury for the elderly. It is estimated that more than 40 million elderly people in my country experience falls every year. Correspondingly, fall detection is an effective passive fall prevention method, which can enable the faller to receive timely medical assistance without human intervention, or activate fall prevention devices (such as inflatable hip protectors) to avoid injuries caused by falls and collisions .

There are many methods of fall detection, which can be divided into pre-collision and post-collision fall detection according to the stage of fall detection completion.

The purpose of pre-collision fall detection is to detect the occurrence of a fall before the body collides with the ground, so as to take timely intervention measures (such as activating the fall prevention device) to prevent human injury. The post-collision fall detection mainly detects the fall event through the impulse generated by the contact between the human body and the ground, ground vibration or sound, and the posture of the human body after the fall, so it cannot provide protection for the human body in the event of a fall collision. Therefore, pre-collision fall detection is more effective than post-collision fall detection.

Pre-collision fall detection mainly judges whether a fall occurs through the kinematics characteristics of the falling process after the body loses balance. Common implementations of pre-collision fall detection include threshold algorithms and machine learning algorithms.

The threshold algorithm usually first determines one or more fall detection indicators (fall detection indicators), usually sports biomechanical indicators, and sets a threshold for them. When the fall detection index exceeds this preset threshold, it means that a fall detection indicator has occurred Fall; otherwise, daily activities (not fall events). Machine learning algorithms, on the other hand, typically use biomechanical data of falls and normal activities as a training set to generate a classification method for classifying falls and daily activities.

Both threshold-based algorithms and machine learning algorithms have some drawbacks. It is difficult to obtain an optimal detection threshold for fall detection based on the threshold algorithm. In the existing threshold algorithm methods, there is a lack of consideration of individual differences. A very low threshold will misjudge non-fall events as fall events (ie, false alarms), while an excessively high threshold will miss real fall events (ie, false alarms). explore). The selection of detection threshold affects the accuracy of fall detection. Machine learning algorithms can overcome the limitations of threshold-based methods to a certain extent, but like threshold-based methods, current machine-learning-based methods cannot reflect individual differences, and individual differences in motion are common.

Contents of the invention

The main purpose of the present invention is to provide a fall detection method and device before a collision, which is used to solve the problem that the existing solutions cannot accurately complete the fall detection and alarm according to individual differences before the human body falls and touches the ground.

The present invention proposes a fall detection method before collision, including:

Determine whether the fall detection index time series is autocorrelated;

If the fall detection index time series is not autocorrelated, use the fall detection index time series of the detected individual to establish a statistical process control model;

If the fall detection index time series is autocorrelated, process the fall detection index time series through the ARIMA model and output non-autocorrelation data;

Establish a statistical process control model based on the non-autocorrelation data processed by the ARIMA model;

Whether the human body falls is judged according to the statistical process control model.

Further, if the fall detection index time series is autocorrelated, the step of processing the fall detection index time series through an ARIMA model and outputting non-autocorrelation data includes,

According to the mixed autoregressive and moving average model ARIMA model, the fall detection index time series x _t is converted into a non-autocorrelated residual value time series e _t according to the following formula: φ,

in,

φ _p represents the regression parameters,

θ _q represents the difference parameter,

B represents the backward shift operator,

p is the autoregressive term,

q is the number of moving average items,

d is the number of differences made when the time series becomes stationary.

Further, according to the mixed autoregressive and moving average model ARIMA (p, d, q) model, the fall detection index time series x _t is converted into a non-autocorrelated residual value time series e _t according to the following formula: After steps include,

Through the maximum likelihood method, the specific values of the φ _p , θ _q , and B are estimated under the premise that the logarithmic likelihood function of the following formula conditions obtains the maximum value:

Among them, the calculation method of S _* (φ _p , θ _q ) is as follows:

in,

n represents the number of items in the fall detection index time series;

T stands for time;

S represents the predicted sum of squares of the time series of residual values.

Further, the step of establishing a statistical process control model based on the non-autocorrelation data processed by the ARIMA model includes,

Based on Shewhart's three-sigma control theory, using the fall detection index time series, the upper/lower control range CL of the statistical process control model is calculated. The control range CL is calculated as follows:

in,

Calculated from the mean of the fall detection index time series of human falls,

Calculated from the moving average of the fall detection index time series of human falls,

_The value of the coefficient c2 is set to 1.128.

Further, the step of judging whether the human body has fallen according to the statistical process control model, and if it is judged to have fallen, issuing a fall alarm includes,

Judging whether the fall detection index time series exceeds the control range CL of the process control chart model;

If it exceeds, it is judged to be in a falling state.

The present invention also proposes a pre-collision fall detection device, comprising:

A verification unit is used to judge whether the fall detection index time series is autocorrelated;

The first execution unit is used to establish a statistical process control model using the fall detection index time series of the detected individual if the fall detection index time series is not autocorrelated;

The second execution unit is used to process the fall detection index time series through the ARIMA model if the time series of the fall detection index is autocorrelated, and output non-autocorrelation data;

The establishment unit is used to establish a statistical process control model according to the non-autocorrelation data processed by the ARIMA model;

The fall alarm unit is used for judging whether the human body has fallen according to the statistical process control model.

Further, the second execution unit includes a first calculation module, which is used to convert the fall detection index time series x _t into non-autocorrelated residual value time according to the following formula according to the mixed autoregressive and moving average model ARIMA model sequence e _t : φ,

in,

φ _p represents the regression parameters,

θ _q represents the difference parameter,

B represents the backward shift operator,

p is the autoregressive term,

q is the number of moving average items,

d is the number of differences made when the time series becomes stationary.

Further, the second execution unit also includes a second calculation module, which is used to estimate the φ _p , θ _q , and B on the premise that the logarithmic likelihood function of the following formula obtains the maximum value through the maximum likelihood method The specific value of:

Among them, the calculation method of S _* (φ _p , θ _q ) is as follows:

in,

n represents the number of items in the fall detection index time series,

T stands for time,

S represents the predicted sum of squares of the time series of residual values.

Further, the establishment unit is used to calculate the upper/lower control range CL of the statistical process control model based on the Shewhart Three Sigma control theory, using the fall detection index time series, and the calculation method of the control range CL is as follows:

in,

Calculated from the mean of the fall detection index time series of human falls,

Calculated from the moving average of the fall detection index time series of human falls,

_The value of the coefficient c2 is set to 1.128.

Further, the fall alarm unit also includes a fall judgment module, which is used to judge whether the fall detection index time series exceeds the control range CL of the process control chart model, and if so, it is judged to be in a fall state.

The beneficial effects of the present invention are: firstly, the scheme establishes a statistical process control chart model for the time series of fall detection indices of different individuals, and uses this model to judge whether the human body activity is a fall, and considers the differences in the actual falls of different users, which can improve The accuracy of fall detection before collision; at the same time, the statistical process control chart model has extremely high timeliness, can quickly detect the occurrence of falls, and provides the possibility for timely intervention measures, such as activating fall prevention devices.

Description of drawings

Fig. 1 is a method flowchart of a collision and fall detection method according to an embodiment of the present invention;

Fig. 2 is a method flowchart of a collision and fall detection method according to another embodiment of the present invention;

3 is a structural block diagram of a collision and fall detection device according to an embodiment of the present invention;

Fig. 4 is a structural block diagram of the second execution unit of the collision and fall detection device according to an embodiment of the present invention.

The realization of the purpose of the present invention, functional characteristics and advantages will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

detailed description

It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relationship between the components in a certain posture (as shown in the accompanying drawings). Relative positional relationship, movement conditions, etc., if the specific posture changes, the directional indication will also change accordingly.

In addition, the descriptions involving "first", "second" and so on in the present invention are only for descriptive purposes, and should not be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In addition, the technical solutions of the various embodiments can be combined with each other, but it must be based on the realization of those skilled in the art. When the combination of technical solutions is contradictory or cannot be realized, it should be considered that the combination of technical solutions does not exist , nor within the scope of protection required by the present invention.

Those skilled in the art will understand that unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the word "comprising" used in the description of the present invention refers to the presence of said features, integers, steps, operations, elements and/or components, but does not exclude the presence or addition of one or more other features, Integers, steps, operations, elements, components, and/or groups thereof. The expression "and/or" used herein includes all or any elements and all combinations of one or more associated listed items.

Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this invention belongs. It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have meanings consistent with their meaning in the context of the prior art, and unless specifically defined as herein, are not intended to be idealized or overly Formal meaning to explain.

ARIMA model: The full name is Autoregressive Integrated Moving Average Model (ARIMA for short). It is a famous time series prediction method proposed by Box and Jenkins in the early 1970s, so it is also called box -jenkins model, Box-Jenkins method. Among them, ARIMA (p, d, q) is called the differential autoregressive moving average model, AR is autoregressive, p is the autoregressive item; MA is the moving average, q is the number of moving average items, and d is the time series when it becomes stable. the number of differences.

The data sequence formed by the forecast object over time is regarded as a random sequence, and a certain mathematical model is used to approximately describe this sequence. This model, once identified, can predict future values from the past and present values of the time series. Modern statistical methods and econometric models have been able to help companies predict the future to some extent.

Statistical process control (SPC for short) is a kind of quality management that applies statistical techniques to evaluate and monitor each stage of the process, establishes and maintains the process at an acceptable and stable level, and ensures that products and services meet the specified requirements. technology.

Likelihood function, in statistics, the likelihood function is a function of the parameters of the statistical model. Given the output x, the likelihood function L(θ|x) with respect to the parameter θ is (numerically) equal to the probability of the variable X given the parameter θ:

L(θ|x)=P(X=x|θ).

The full name of CL is Control Limit, and the Chinese name is control range.

Referring to FIG. 1 , an embodiment of the present invention is proposed, a method for detecting a fall before a collision, comprising the following steps:

S10. Determine whether the fall detection index time series is autocorrelated;

S11. If the fall detection index time series is not autocorrelated, use the fall detection index time series of the detected individual to establish a statistical process control model;

S12. If the fall detection index time series is autocorrelated, process the fall detection index time series through the ARIMA model, and output non-autocorrelation data;

S13. Establish a statistical process control model according to the non-autocorrelation data processed by the ARIMA model;

S14. Determine whether the human body has fallen according to the statistical process control model.

For step S10, one assumption for effective use of statistical process control models is that the time series data be independent. This assumption is often not met in applications due to the autocorrelation of the data. Therefore, before applying the statistical process control model, it is necessary to test the autocorrelation of the fall detection index time series, specifically through the method of the autocorrelation coefficient equation diagram.

For step S11, the autocorrelation of the fall detection index time series is determined by the method of the autocorrelation coefficient equation graph. If the fall detection index time series is not autocorrelated, the statistical process control model will be established directly by the test fall detection index time series.

For step S12, determine the autocorrelation of the fall detection index time series by the method of the autocorrelation coefficient equation diagram, if the fall detection index time series is autocorrelated, because an assumption of the effective use of the statistical process control model is that the time series data should be independent, Therefore, it is not possible to directly use the fall detection index time series to establish a statistical process control model. It is necessary to process the fall detection index time series through the ARIMA model, and finally output non-autocorrelation data that can be used to establish a statistical process control model.

For step S13, according to the non-autocorrelation data processed by the ARIMA model, a statistical process control model is established on the basis of the time series of fall detection index of the detected individual, so that even if the time series of fall detection index is autocorrelated at the beginning, it can eventually be Used to build statistical process control models.

For step S14, judge whether the specified individual has fallen according to the above-mentioned statistical process control model, and when the fall detection index time series exceeds the control range of the statistical process control chart model, it is determined that the specified individual has fallen, and further intervention measures can be activated, and an alarm is issued at the same time .

A pre-collision fall detection method of the present invention first establishes a statistical process control chart model for the fall detection index time series of different individuals, and uses this model to judge whether human body activity is a fall, taking into account the differences in the actual falls of different users, which can be used Improve the accuracy of fall detection before collision; at the same time, the statistical process control chart model has extremely high timeliness, can quickly detect the occurrence of falls, and provides the possibility for timely intervention measures, such as activating fall prevention devices.

Referring to FIG. 2 , an embodiment of the present invention is proposed, a method for detecting a fall before a collision, including the following steps:

S20. Determine whether the fall detection index time series is autocorrelated;

S21. If the fall detection index time series is not autocorrelated, use the fall detection index time series of the detected individual to establish a statistical process control model;

S22. If the fall detection index time series is autocorrelated, process the fall detection index time series through the ARIMA model, and output non-autocorrelation data;

S23. Establish a statistical process control model according to the non-autocorrelation data processed by the ARIMA model;

S24. Judging whether the time series of the fall detection index exceeds the control range of the process control chart model, and if so, it is judged as a fall state.

For step S20, one assumption for effective use of statistical process control models is that the time series data be independent. This assumption is often not met in applications due to the autocorrelation of the data. Therefore, before applying the statistical process control model, it is necessary to test the autocorrelation of the fall detection index time series, specifically through the method of the autocorrelation coefficient equation diagram.

For step S21, the autocorrelation of the fall detection index time series is determined by the method of the autocorrelation coefficient equation diagram. If the fall detection index time series is not autocorrelated, the statistical process control model will be established directly by the test fall detection index time series.

For step S22, determine the autocorrelation of the test fall detection index time series by the method of the autocorrelation coefficient equation graph, if the fall detection index time series is autocorrelated, because an assumption of the effective use of the statistical process control model is that the time series data should be independent, Therefore, it is not possible to directly use the fall detection index time series to establish a statistical process control model. It is necessary to process the fall detection index time series through the ARIMA model, and finally output non-autocorrelation data that can be used to establish a statistical process control model.

The specific step S22 includes the following steps:

S221. According to the mixed autoregressive and moving average model ARIMA (p, d, q) model, the fall detection index time series x _t is converted into a non-autocorrelated residual value time series e _t according to the following formula: φ,

Among them, φ _p represents the regression parameter, θ _q represents the difference parameter, B represents the backward shift operator, p is the autoregressive item, q is the number of moving average items, and d is the number of differences made when the time series becomes stationary.

S222. Estimate the specific values of φ _p , θ _q , and B by using the maximum likelihood method and ensuring that the logarithmic likelihood function of the following formula obtains the maximum value:

Among them, the calculation method of S _* (φ _p , θ _q ) is as follows:

in,

n represents the number of items in the fall detection index time series,

T stands for time,

S represents the predicted sum of squares of the time series of residual values.

Specifically, through steps S221 and S222, the autocorrelated fall detection index time series x _t is further processed to obtain non-autocorrelated data suitable for establishing a statistical process control model.

Specifically, step S23 is: based on the Shewhart Three Sigma control theory, using the fall detection index time series to calculate the upper/lower control range CL of the statistical process control model, the calculation method of the control range CL is as follows:

in,

Calculated from the mean of the fall detection index time series of human falls,

Calculated from the moving average of the fall detection index time series of human falls,

_The value of the coefficient c2 is set to 1.128.

For step S23, according to the non-autocorrelation data processed by the ARIMA model, a statistical process control model is established based on the time series of fall detection index of the detected individual, so that even if the time series of fall detection index is autocorrelated at the beginning, it can eventually Used to build statistical process control models.

For step S24, by judging whether the fall detection index time series exceeds the control range CL of the process control chart model, if it exceeds, it is determined that the specified individual has fallen, and further intervention measures can be activated, and an alarm is issued at the same time.

In a specific application example of the present invention, an elderly person accidentally slips and falls while walking. Using the solution of the present invention, the fall event can be judged within 120 ms after the initial slip. After the body collides with the ground, relying on the advance of the fall detection, the airbag and other fall prevention devices can be activated (the activation time is reported to be about 70ms), thereby preventing the occurrence of injuries.

A pre-collision fall detection method of the present invention first establishes a statistical process control chart model for the fall detection index time series of different individuals, and uses this model to judge whether human body activity is a fall, taking into account the differences in the actual falls of different users, which can be used Improve the accuracy of fall detection before collision; at the same time, the statistical process control chart model has extremely high timeliness, can quickly detect the occurrence of falls, and provides the possibility for timely intervention measures, such as activating fall prevention devices.

Referring to Fig. 3 and Fig. 4, another embodiment of the present invention is proposed, a pre-collision fall detection device, comprising:

The checking unit 10 is used for judging whether the fall detection index time series is autocorrelated.

The first execution unit 30 is configured to use the time series of fall detection indices of the detected individual to establish a statistical process control model if the time series of fall detection indices is not autocorrelated.

The second execution unit 20 is configured to process the fall detection index time series through an ARIMA model if the fall detection index time series is autocorrelated, and output non-autocorrelation data.

The establishment unit 40 is configured to establish a statistical process control model according to the non-autocorrelation data processed by the ARIMA model.

The fall alarm unit 50 is used for judging whether the human body has fallen according to the statistical process control model.

For the verification unit 10 , an assumption for effective use of the statistical process control model is that the time series data should be independent, which is often not satisfied in applications due to the autocorrelation of the data. Therefore, before applying the statistical process control model, it is necessary to test the autocorrelation of the fall detection index time series, specifically through the method of the autocorrelation coefficient equation diagram.

For the first execution unit 30, the autocorrelation of the time series of the test fall detection index is determined by the method of the autocorrelation coefficient equation diagram. If the time series of the fall detection index is not autocorrelated, the statistical process will be established by directly using the time series of the test fall detection index control model.

For the second execution unit 20, the autocorrelation of the test fall detection index time series is determined by the method of the autocorrelation coefficient equation graph, if the fall detection index time series is autocorrelated, because one assumption for effective use of statistical process control models is time series data To be independent, it is not possible to directly use the fall detection index time series to establish a statistical process control model. It is necessary to process the fall detection index time series through the ARIMA model, and finally output non-autocorrelation data that can be used to establish a statistical process control model.

The second execution unit 20 includes a first calculation module 21, which is used to convert the fall detection index time series x _t into non-autocorrelated residuals according to the following formula according to the mixed autoregressive and moving average model ARIMA (p, d, q) model difference time series e _t : φ,

in,

φ _p represents the regression parameters,

θ _q represents the difference parameter,

B represents the backward shift operator.

The second execution unit 20 also includes a second calculation module 22, which is used for estimating the specific values of φ _p , θ _q , and B under the premise that the logarithmic likelihood function of the following formula is guaranteed to obtain the maximum value through the maximum likelihood method :

Among them, the calculation method of S _* (φ _p , θ _q ) is as follows:

in,

n represents the number of items in the fall detection index time series;

T stands for time,

S represents the predicted sum of squares of the time series of residual values.

Specifically, through the first calculation module 21 and the second calculation module 22, the autocorrelated fall detection index time series x _t is further processed to obtain non-autocorrelation data suitable for establishing a statistical process control model.

Specifically, the establishment unit 40 is used to calculate the upper/lower control range CL of the statistical process control model based on Shewhart's three-sigma control theory and using the fall detection index time series. The calculation method of the control range CL is as follows:

in,

Calculated from the mean of the fall detection index time series of human falls,

Calculated from the moving average of the fall detection index time series of human falls, the value of the coefficient c2 is set to _1.128 .

For the establishment unit 40, according to the non-autocorrelation data processed by the ARIMA model, the statistical process control model is established on the basis of the time series of the detected individual's fall detection index, so that even if the time series of the fall detection index is autocorrelated at the beginning, it will eventually be Can be used to build statistical process control models.

Specifically, the fall alarm unit 50 also includes a fall judgment module for judging whether the fall detection index time series exceeds the control range CL of the process control chart model, and if so, a fall alarm is issued.

For the fall judgment module, by judging whether the fall detection index time series exceeds the control range CL of the process control chart model, if it exceeds, it is determined that the specified individual has fallen, and further intervention measures can be activated, and an alarm is issued at the same time.

A pre-collision fall detection device of the present invention first establishes a statistical process control chart model for the time series of fall detection indices of different individuals, and uses this model to judge whether human body activity is a fall, taking into account the differences in the actual falls of different users, It can improve the accuracy of fall detection before collision; at the same time, the statistical process control chart model has extremely high timeliness, can quickly detect the occurrence of falls, and provides the possibility for timely intervention measures, such as activating fall prevention devices.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process conversion made by using the description of the present invention and the contents of the accompanying drawings, or directly or indirectly used in other related All technical fields are equally included in the scope of patent protection of the present invention.

Claims (10)

1. A pre-collision fall detection method, characterized in that, comprising, Determine whether the fall detection index time series is autocorrelated; If the fall detection index time series is not autocorrelated, use the fall detection index time series of the detected individual to establish a statistical process control model; If the fall detection index time series is autocorrelated, process the fall detection index time series through the ARIMA model and output non-autocorrelation data; Establish a statistical process control model based on the non-autocorrelation data processed by the ARIMA model; Whether the human body falls is judged according to the statistical process control model. 2. The fall detection method before the collision as claimed in claim 1, wherein, if the fall detection index time series is autocorrelated, the fall detection index time series is processed by the ARIMA model, and the non-autocorrelation data step is output, comprising , According to the mixed autoregressive and moving average model ARIMA model, the fall detection index time series x _t is converted into a non-autocorrelated residual value time series e _t according to the following formula: φ, <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>p</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>p</mi> </msubsup> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <msup> <mi>B</mi> <mi>p</mi> </msup> <mo>)</mo> <msup> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>B</mi> <mo>)</mo> </mrow> <mi>d</mi> </msup> <msub> <mi>x</mi> <mi>t</mi> </msub> <mo>=</mo> <mo>(</mo> <mn>1</mn> <mo>+</mo> <msubsup> <mi>&amp;Sigma;</mi> <mi>q</mi> <mi>q</mi> </msubsup> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <msup> <mi>B</mi> <mi>q</mi> </msup> <mo>)</mo> <msub> <mi>e</mi> <mi>t</mi> </msub> <mo>,</mo> </mrow> in, φ _p represents the regression parameters, θ _q represents the difference parameter, B represents the backward shift operator, p is the autoregressive term, q is the number of moving average items, d is the number of differences made when the time series becomes stationary. 3. fall detection method before collision as claimed in claim 2 is characterized in that, described according to mixed autoregressive and moving average model ARIMA (p, d, q) model, to fall detection index time series x _t according to following formula Convert to a non-autocorrelated time series of residual values e _t : After steps include, Through the maximum likelihood method, the specific values of the φ _p , θ _q , and B are estimated under the premise that the logarithmic likelihood function of the following formula conditions obtains the maximum value: <mrow> <mi>l</mi> <mo>*</mo> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>,</mo> <msubsup> <mi>&amp;sigma;</mi> <msub> <mi>e</mi> <mi>t</mi> </msub> <mn>2</mn> </msubsup> <mo>)</mo> </mrow> <mo>=</mo> <mo>-</mo> <mfrac> <mi>n</mi> <mn>2</mn> </mfrac> <mi>l</mi> <mi>n</mi> <mrow> <mo>(</mo> <msubsup> <mi>&amp;sigma;</mi> <msub> <mi>e</mi> <mi>t</mi> </msub> <mn>2</mn> </msubsup> <mo>)</mo> </mrow> <mo>-</mo> <mfrac> <mrow> <mi>S</mi> <mo>*</mo> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <mn>2</mn> <msubsup> <mi>&amp;sigma;</mi> <msub> <mi>e</mi> <mi>t</mi> </msub> <mn>2</mn> </msubsup> </mrow> </mfrac> <mo>,</mo> </mrow> Among them, the calculation method of S*(φ _p ,θ _q ) is as follows: <mrow> <mi>S</mi> <mo>*</mo> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>T</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </msubsup> <msubsup> <mi>e</mi> <mi>t</mi> <mn>2</mn> </msubsup> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>|</mo> <mo>(</mo> <mrow> <mn>1</mn> <mo>-</mo> <mi>B</mi> </mrow> <mo>)</mo> <msub> <mi>x</mi> <mrow> <mi>t</mi> <mo>*</mo> </mrow> </msub> <mo>,</mo> <msub> <mi>e</mi> <mrow> <mi>t</mi> <mo>*</mo> </mrow> </msub> <mo>,</mo> <mo>(</mo> <mrow> <mn>1</mn> <mo>-</mo> <mi>B</mi> </mrow> <mo>)</mo> <msub> <mi>x</mi> <mi>t</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> </mrow> in, n represents the number of items in the fall detection index time series, T stands for time, S represents the predicted sum of squares of the time series of residual values. 4. The fall detection method before collision as claimed in claim 3, wherein the step of establishing a statistical process control model according to the non-autocorrelation data processed by the ARIMA model comprises, Based on Shewhart's three-sigma control theory, using the fall detection index time series, the upper/lower control range CL of the statistical process control model is calculated. The control range CL is calculated as follows: <mrow> <msub> <mi>CL</mi> <mrow> <mo>(</mo> <mi>U</mi> <mi>p</mi> <mi>p</mi> <mi>e</mi> <mi>r</mi> <mo>|</mo> <mi>L</mi> <mi>o</mi> <mi>w</mi> <mi>e</mi> <mi>r</mi> <mo>)</mo> </mrow> </msub> <mo>=</mo> <mover> <mi>X</mi> <mo>&amp;OverBar;</mo> </mover> <munder> <mo>+</mo> <mo>&amp;OverBar;</mo> </munder> <mn>3</mn> <mi>s</mi> <mi>i</mi> <mi>g</mi> <mi>m</mi> <mi>a</mi> <mo>=</mo> <mover> <mi>X</mi> <mo>&amp;OverBar;</mo> </mover> <munder> <mo>+</mo> <mo>&amp;OverBar;</mo> </munder> <mn>3</mn> <mfrac> <mover> <mrow> <mi>m</mi> <mi>r</mi> </mrow> <mo>&amp;OverBar;</mo> </mover> <msub> <mi>c</mi> <mn>2</mn> </msub> </mfrac> <mo>,</mo> </mrow> in, Calculated from the mean of the fall detection index time series of human falls, Calculated from the moving average of the fall detection index time series of human falls, the value of the coefficient c2 is set to _1.128 . 5. The fall detection method before collision as claimed in claim 4, wherein the step of judging whether the human body falls according to the statistical process control model comprises, Judging whether the fall detection index time series exceeds the control range CL of the process control chart model; If it exceeds, it is judged to be in a falling state. 6. A pre-collision fall detection device, characterized in that it comprises, The verification unit is used to judge whether the fall detection index time series is autocorrelated; The first execution unit is used to establish a statistical process control model using the fall detection index time series of the detected individual if the fall detection index time series is not autocorrelated; The second execution unit is used to process the fall detection index time series through the ARIMA model if the time series of the fall detection index is autocorrelated, and output non-autocorrelation data; The establishment unit is used to establish a statistical process control model according to the non-autocorrelation data processed by the ARIMA model; The fall alarm unit is used for judging whether the human body has fallen according to the statistical process control model. 7. The pre-collision fall detection device according to claim 6, characterized in that, the second execution unit includes a first calculation module, which is used to perform a fall detection index time according to a hybrid autoregressive and moving average model ARIMA model The sequence x _t is transformed into a non-autocorrelated residual value time series e _t according to the following formula: φ, <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>p</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>p</mi> </msubsup> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <msup> <mi>B</mi> <mi>p</mi> </msup> <mo>)</mo> <msup> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>B</mi> <mo>)</mo> </mrow> <mi>d</mi> </msup> <msub> <mi>x</mi> <mi>t</mi> </msub> <mo>=</mo> <mo>(</mo> <mn>1</mn> <mo>+</mo> <msubsup> <mi>&amp;Sigma;</mi> <mi>q</mi> <mi>q</mi> </msubsup> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <msup> <mi>B</mi> <mi>q</mi> </msup> <mo>)</mo> <msub> <mi>e</mi> <mi>t</mi> </msub> <mo>,</mo> </mrow> in, φ _p represents the regression parameters, θ _q represents the difference parameter, B represents the backward shift operator, p is the autoregressive term, q is the number of moving average items, d is the number of differences made when the time series becomes stationary. 8. The pre-collision fall detection device according to claim 7, wherein the second execution unit further includes a second calculation module, which is used to ensure that the conditional logarithmic likelihood function of the following formula is obtained by the maximum likelihood method Under the premise of the maximum value, estimate the specific values of φ _p , θ _q , and B: <mrow> <mn>1</mn> <mo>*</mo> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>,</mo> <msubsup> <mi>&amp;sigma;</mi> <msub> <mi>e</mi> <mi>t</mi> </msub> <mn>2</mn> </msubsup> <mo>)</mo> </mrow> <mo>=</mo> <mo>-</mo> <mfrac> <mi>n</mi> <mn>2</mn> </mfrac> <mi>l</mi> <mi>n</mi> <mrow> <mo>(</mo> <msubsup> <mi>&amp;sigma;</mi> <msub> <mi>e</mi> <mi>t</mi> </msub> <mn>2</mn> </msubsup> <mo>)</mo> </mrow> <mo>-</mo> <mfrac> <mrow> <mi>S</mi> <mo>*</mo> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <mn>2</mn> <msubsup> <mi>&amp;sigma;</mi> <msub> <mi>e</mi> <mi>t</mi> </msub> <mn>2</mn> </msubsup> </mrow> </mfrac> <mo>,</mo> </mrow> Among them, the calculation method of S*(φ _p ,θ _q ) is as follows: <mrow> <mi>S</mi> <mo>*</mo> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <msubsup> <mi>&amp;Sigma;</mi> <mrow> <mi>T</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </msubsup> <msubsup> <mi>e</mi> <mi>t</mi> <mn>2</mn> </msubsup> <mrow> <mo>(</mo> <msub> <mi>&amp;phi;</mi> <mi>p</mi> </msub> <mo>,</mo> <msub> <mi>&amp;theta;</mi> <mi>q</mi> </msub> <mo>|</mo> <mo>(</mo> <mrow> <mn>1</mn> <mo>-</mo> <mi>B</mi> </mrow> <mo>)</mo> <msub> <mi>x</mi> <mrow> <mi>t</mi> <mo>*</mo> </mrow> </msub> <mo>,</mo> <msub> <mi>e</mi> <mrow> <mi>t</mi> <mo>*</mo> </mrow> </msub> <mo>,</mo> <mo>(</mo> <mrow> <mn>1</mn> <mo>-</mo> <mi>B</mi> </mrow> <mo>)</mo> <msub> <mi>x</mi> <mi>t</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> </mrow> 2 in, n represents the number of items in the fall detection index time series; T stands for time; S represents the predicted sum of squares of the time series of residual values. 9. The pre-collision fall detection device according to claim 8, wherein the establishment unit is used to calculate the upper/lower value of the statistical process control model based on the Shewhart Three Sigma control theory, using the fall detection index time series The lower control range CL, the calculation method of the control range CL is as follows: <mrow> <msub> <mi>CL</mi> <mrow> <mo>(</mo> <mi>U</mi> <mi>p</mi> <mi>p</mi> <mi>e</mi> <mi>r</mi> <mo>|</mo> <mi>L</mi> <mi>o</mi> <mi>w</mi> <mi>e</mi> <mi>r</mi> <mo>)</mo> </mrow> </msub> <mo>=</mo> <mover> <mi>X</mi> <mo>&amp;OverBar;</mo> </mover> <munder> <mo>+</mo> <mo>&amp;OverBar;</mo> </munder> <mn>3</mn> <mi>s</mi> <mi>i</mi> <mi>g</mi> <mi>m</mi> <mi>a</mi> <mo>=</mo> <mover> <mi>X</mi> <mo>&amp;OverBar;</mo> </mover> <munder> <mo>+</mo> <mo>&amp;OverBar;</mo> </munder> <mn>3</mn> <mfrac> <mover> <mrow> <mi>m</mi> <mi>r</mi> </mrow> <mo>&amp;OverBar;</mo> </mover> <msub> <mi>c</mi> <mn>2</mn> </msub> </mfrac> <mo>,</mo> </mrow> in, Calculated from the mean of the fall detection index time series of human falls, Calculated from the moving average of the fall detection index time series of human falls, the value of the coefficient c2 is set to _1.128 . 10. The pre-collision fall detection device according to claim 9, wherein the fall alarm unit also includes a fall judgment module for judging whether the fall detection index time series exceeds the control range CL of the process control diagram model, If it exceeds, it is judged to be in a falling state.