WO2018161433A1 - Wireless signal transmission-based indoor fire detection and alarm method and system - Google Patents

Abstract

The present invention provides a wireless signal transmission-based indoor fire detection and alarm method and system; said wireless signal transmission-based indoor fire detection and alarm method comprises the following steps: step S1: a receiving terminal receiving an RF signal from a transmitting terminal, obtaining CSI data, then using a Butterworth low-pass filter to perform Gaussian white-noise removal processing on the data, and waiting until the current environment enters a stable state; step S2: performing statistical probability on the data of the current environment in a stable state, and calculating the angle of arrival, amplitude feature, and phase diversity of same; step 3: continuing to acquire CSI data of the current environment, and calculating to obtain the overall probability of a fire occurring. In the present invention, an existing Wi-Fi device is used to collect signals so as to accomplish the subsequent processing and analysis of the data; thus it is possible to make a valid determination of a current indoor environment without requiring additional devices and calibration, reducing overhead; the invention is easy to use and is widely applicable.

Description

Indoor fire detection and alarm method and system based on wireless signal transmission Technical field

The invention relates to an indoor fire detection method, in particular to a method for indoor fire detection and alarm based on wireless signal transmission, and relates to a system for indoor fire detection and alarm based on wireless signal transmission.

Background technique

A fire is a disaster caused by burning that is out of control in time and space. In today's society, fire is one of the most important disasters that threaten social public safety and endanger people's lives and property. It is also a disaster with multiple frequencies and the largest time and space in multiple disasters. From the data of fire accidents in recent years, the number of casualties caused by fires in China is second only to mine disasters. The number of fires per year is close to 400,000, and the number of deaths is as high as 1,800. The direct economic losses caused by fires amount to 4 billion yuan. Down, there are more than 1,000 fires every day, and these data are showing an increasing trend.

Major fire accidents in recent years include: First, on August 9, 2015, 165 people died in the dangerous goods warehouse of Tianjin Binhai New Area Ruihai International Logistics Co., Ltd., and 8 people were missing. Second, February 5, 2015, Huizhou City, Guangdong Province 17 people died in Huidong County Compulsory Commodity City; third, on November 15, 2010, 58 people were killed in high-rise residential buildings on Jiaozhou Road, Yuyao Road, Shanghai; fourth, on November 5, 2010, 19 people died in Jilin Commercial Building, and so on.

This bloody case reveals to us the great danger of fire, but also can not cause us to think. If we can find the fire in time in the early stage of the fire, report the fire, it can be carried out in the shortest time. Fight to minimize the damage caused by fire. Therefore, fire monitoring is a vital part of the entire fire prevention and control system.

Fire detection relies on a method that can detect fire information early. For now, it relies mainly on a series of dedicated fire detection devices, such as smoke detectors commonly found in public places. However, it is not practical to require special fire detection equipment in all places due to social and economic conditions and people's awareness of disaster prevention and mitigation. For example, in China, most households do not have corresponding fire monitoring equipment installed. However, considering the development of the Internet and Wi-Fi technology, the popularity of commercial Wi-Fi continues to climb. Inspired by this, we consider whether we can use existing commercial Wi-Fi equipment to achieve timely fire detection and early warning, thus helping to reduce fire damage and protect the lives and property of users. To this end, in this paper, we propose to use existing Wi-Fi equipment to research and design a Wi-Fi-based indoor fire monitoring system without the need to install additional dedicated equipment. It can monitor the indoor environment in real time. If there is a fire, the system can take timely measures, such as issuing an alarm signal, notifying the owner or alarm. The system uses existing general-purpose commercial Wi-Fi equipment with low cost advantages and universal applicability and versatility.

Summary of the invention

In order to overcome the disadvantage that the special smoke detector is not popular in China, the present invention provides a low cost and high precision method and system for indoor fire detection and alarm based on wireless network signals (radio frequency signals), aiming to specify In the indoor environment, by using the existing wireless network and equipment, the indoor environment can be effectively detected without prior training, and if an abnormality occurs, the timely alarm and feedback can be achieved.

In this regard, the present invention provides a method for indoor fire detection and alarm based on wireless signal transmission, comprising the following steps:

In step S1, the receiving end receives the radio frequency signal from the transmitting end, obtains the CSI data, uses the Butterworth low-pass filter to remove the Gaussian white noise, and determines whether there is human or non-stabilizing factor interference in the current environment by using the variance method. Then, it is determined that the current environment is in an unstable state, waiting for the current environment to enter a steady state, if otherwise, directly jump to step S2;

Step S2, respectively performing statistical probability on the current environment as a steady state data, and calculating an angle of arrival, an amplitude characteristic, and a phase difference characteristic;

Step S3, continuously collecting the CSI data of the current environment, calculating the angle of arrival, and _{obtaining the} probability of occurrence of the fire in the current environment P _codition1 , and comparing the probability of the current environment with the steady state to obtain the fire probability P _{codition2 of the} current environment. The current environmental fire probability P _{codition3 is} calculated by randomness judgment. Finally, the total fire occurrence probability P _fire =P _codition1 +P _codition2 +P _{codition3 is} calculated by calculation to determine whether the room is on fire.

A further improvement of the invention is that said step S1 comprises the following substeps:

Step S11, collecting CSI data in a unit time period in the current environment, and obtaining a matrix of the number of links × the number of subcarriers;

Step S12, performing low-pass filtering processing on the CSI data by using a Butterworth low-pass filter to remove the influence of the white Gaussian noise;

Step S13, determining whether the mean square error is within the first threshold range by calculating the mean and the mean square error of each subcarrier in the unit time period, and if the first threshold value range is exceeded, determining whether the current environment has a person's activity or an unstable factor Interference, wait until the person leaves or the current environment returns to a stable state, and then jumps to step S2.

A further improvement of the invention is that said step S2 comprises the following substeps:

Step S21, extracting CSI data in a time period T0. In this time period T0, the data of each subcarrier is first converged, then the number of occurrences of the data is counted, and the number of times the data combination occurs in sequence Statistics are performed to calculate the cumulative frequency of each data and data combination. The probability of occurrence of n different data and the probability of occurrence of m different data combinations are obtained by the cumulative frequency, and the state space probability and joint probability are obtained. Finally, the combination of interference data and interference data with a probability less than or equal to 1% is removed as the state space probability P={P _X(0) ,...,P _{X(i) of the} current environment in a steady state. ...} and joint probability Q={Q _Y(0) ,...,Q _Y(i) ,...}, and record;

Step S22, using the MUSIC algorithm, calculating the ratio between the phase difference and the antenna spacing by using the phase difference between the same subcarrier and the adjacent antenna, and obtaining the arrival angle of the subcarrier, thereby using the current environment as the steady state. Arrival angle and record;

Step S23, recording amplitude characteristics and phase difference characteristics of the current environment in a steady state;

Step S24, determining whether the mean square error is within the first threshold range by calculating the mean value and the mean square error of each subcarrier in the time period T0. If the first threshold value range is exceeded, determining whether the current environment has a person's activity or an unstable factor Interference, wait until the person leaves or the current environment returns to a stable state, and then jumps to step S3.

A further improvement of the invention is that said step S3 comprises the following substeps:

Step S31, extracting CSI data in a unit time T1, wherein T1<<T0, determining whether the mean square error is within the first threshold range by calculating the mean and the mean square error of each subcarrier in the unit time T1, if The first threshold range, it is determined that the current environment exists human activity or non-stability factor interference, wait until the person leaves or the current environment returns to a stable state, then jumps to step S32;

Step S32, loading tolerance

Using the MUSIC algorithm, the phase difference between the antenna and the antenna is calculated by the phase difference between the adjacent subcarriers, and the arrival angle γ of the subcarrier is calculated.

in case

The probability of a fire in the current environment is

in case

Then the fire occurrence probability of the current environment is P _fire =P _codition1 =T _codition1 ; wherein γ ₀ is the average of the arrival angles in the steady state,

To tolerate the range of angular deviation, T _codition1 is the _weight for judging the possibility of occurrence of fire by calculating the angle of arrival; the probability of occurrence of the fire is also called the probability of fire;

Step S33, extracting CSI data in unit time T1, in this unit time T1, first converging data of each subcarrier, then counting the number of occurrences of the data, and performing the number of occurrences of the data combinations appearing in sequence Statistics, calculate the cumulative frequency of each data and data combination, and obtain the probability of occurrence of n' different data and the probability of occurrence of data combination with different m' by cumulative frequency, and calculate the state space probability and joint probability. Finally, the combination of interference data and interference data with a probability less than or equal to 1% is removed as the state space probability P'={P' _X(0) ,...,P' _{X of the} current environment in a steady state. _(i) ,...} and joint probability Q'={Q' _Y(0) ,...,Q' _Y(i) ,...};

Step S34, using a chi-square test method, calculating

with

N is the number of elements different from X(i) in P and P', M is the number of elements different from Y(i) in Q and Q', and X(s) is the smallest element in P and P' The element, X(e) is the largest element of the elements in P and P', Y(s) is the element with the smallest element in Q and Q', and Y(e) is the element with the largest element in Q and Q'; The current chi-square test uses the confidence interval α _{1 to} compare the sizes of χ ₁ ² and χ ² (N-1) and χ ₂ ² and χ ² (M-1) by looking up the table, if and only if χ ₁ ² >χ ² (N-1) and χ ₂ ² >χ ² (M-1) is established, proceed to step S35, otherwise,

Calculating _{P fire} = P codition1 + P _codition2 is greater than Δ _ALARM, is greater than the alarm, otherwise, returns to step S31 to continue to monitor; _wherein, T codition2 is a weight comparison determination fire probability by the probability that the environment and the steady state current _weight, T codition3 In order to determine the weight of the fire probability by randomness, _ΔALARM is the threshold of whether or not to alarm;

Step S35, checking the randomness, using the run-length test method, dividing the elements in the two difference sets P-P' and Q-Q' into two parts according to the frequency average, so that the two parts of the frequency are the same and the part is the same. The element is 1, and the other part is 0, which is substituted into the CSI data sequence, and the sequence containing only 0 and 1 is extracted. The number of consecutive occurrences of the same value in the sequence is recorded as the run number r, and the number of 1 will appear. Recorded as n ₁ , the number of occurrences of 0 is recorded as n ₂ , and the mean value of the sampling distribution is calculated.

And sampling distribution variance

Calculate run statistics

The confidence interval α ₂ used for loading the current run test is used to perform a table lookup using the run statistic to obtain the current probability p, and it is judged whether or not p<1-α ₂ is satisfied. If not, the randomness is not satisfied, and the process returns to step S31, if it is satisfied. , it is established that random, then _P codition3 = _T codition4, calculate _{P fire = P codition1 + P codition2} + P codition3 is greater than Δ _aLARM, is greater than the alarm, otherwise, returns to step S31 to continue to monitor; _wherein, T codition4 through The probability of the fire probability is judged by comparing the probability of the current environment with the steady state;

In step S36, a fire alarm signal is issued. If the alarm is not turned off in time, a message is sent to the owner. If the owner determines the alarm as a false alarm, the parameter is traversed by the feasible value of the previous parameter and the current CSI data. T _codition1 , T _codition2 , T _codition3 , T _{codition4 ,} and Δ _ALARM are corrected.

A further improvement of the present invention is that the number of the transmitting ends is 1, and the number of the receiving ends is three or more.

The invention also provides a system for indoor fire detection and alarm based on wireless signal transmission, comprising:

a CSI data acquisition module, configured to receive a radio frequency signal from the transmitting end, calculate CSI data, and remove a Gaussian white noise by using a Butterworth low-pass filter;

The data processing module separately performs statistical probability on the data, and calculates its angle of arrival, chi-square test and run-length test;

The fire condition judging module calculates the possibility of the above data, and determines whether the current environment has a fire by whether it is in the second threshold range;

An alarm module, configured to issue a fire alarm signal and notify the owner when a fire occurs in the current environment;

The feedback correction module is used to correct the parameters when the alarm is determined to be a false alarm.

According to a further improvement of the present invention, the CSI data acquisition module includes:

The sensing unit is configured to initialize the channel state data to obtain a matrix of the number of links×the number of subcarriers; and the filtering unit uses the Butterworth algorithm to remove the influence of the Gaussian white noise on the channel state data.

According to a further improvement of the present invention, the data processing module comprises:

a data feature calculation unit, configured to extract features for CSI data, the extracted features include performing mean square error determination, statistical probability, calculating an angle of arrival, calculating a chi-square test, and calculating a run-length test on the CSI data;

The first abnormal output unit performs a calculation on the probability of interference of the human activity on the data, and determines whether it is within a limited first threshold range, thereby determining whether there is interference of the human activity or the unsteady factor in the room;

The second abnormality output unit calculates the fire possibility of the data, and determines whether there is a fire in the room by whether it is within a limited second threshold range.

A further improvement of the present invention is that, in the alarm module, after the fire alarm signal is issued and the homeowner is notified, if the alarm is not responded or eliminated in time, a help signal is sent to the police.

A further improvement of the present invention is that in the feedback correction module, all feasible solutions of relevant parameters are saved, the values of the relevant parameters are adjusted by the current CSI data, and the combination of feasible solutions is reduced to increase the accuracy. Related parameters include T _codition1 , T _codition2 , T _codition3 , T _{codition4 ,} and Δ _ALARM .

Compared with the prior art, the invention has the beneficial effects that the existing WIFI device can be used to collect signals, thereby implementing subsequent processing and analysis of the data, and realizing the current environment in the room without prior training. Effective judgment, no need to install additional equipment, saves overhead, and has a popular type; on the basis of this, the invention is convenient to use, requires no additional calibration, and has universal applicability.

DRAWINGS

1 is a schematic diagram of a workflow of an embodiment of the present invention;

2 is a first schematic diagram showing the effect of fire on a signal in an embodiment of the present invention;

3 is a second schematic diagram showing the effect of fire on a signal in an embodiment of the present invention;

4 is a schematic diagram of a data processing principle according to an embodiment of the present invention;

FIG. 5 is a system architecture diagram of an embodiment of the present invention; FIG.

Figure 6 is a flow chart of an indoor fire alarm according to an embodiment of the present invention.

detailed description

The preferred embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

As shown in FIG. 1, this example provides a method for indoor fire detection and alarm based on wireless signal transmission, including the following steps:

In step S1, the receiving end receives the radio frequency signal from the transmitting end, and obtains the CSI data and uses the Butterworth low pass. The filter removes the Gaussian white noise processing data, and determines whether there is human or non-stabilizing factor interference in the current environment by using the variance method. If yes, it determines that the current environment is in an unstable state, waiting for the current environment to enter a stable state, if otherwise, directly jumps to Step S2;

Step S2, respectively performing statistical probability on the current environment as a steady state data, and calculating an angle of arrival, an amplitude characteristic, and a phase difference characteristic;

Step S3, continuously collecting the CSI data of the current environment, calculating the angle of arrival, and _{obtaining the} probability of occurrence of the fire in the current environment P _codition1 , and comparing the probability of the current environment with the steady state to obtain the fire probability P _{codition2 of the} current environment. The current environmental fire probability P _{codition3 is} calculated by randomness judgment. Finally, the total fire occurrence probability P _fire =P _codition1 +P _codition2 +P _{codition3 is} calculated by calculation to determine whether the room is on fire.

In the step S3, if there is a fire, a fire alarm signal is issued and the homeowner is notified, and if the alarm is not responded or eliminated in time, the police will be asked for help; if the alarm is considered to be misjudged, the parameters are corrected.

In practical applications, we use the Intel 5300 wireless network card as the receiving end to receive data, and the transmitting end is the wireless router AP. This example is based on the propagation of indoor radio frequency signals. Because of the fire, the flame will change the indoor environment, which will cause the signal propagation path to change, as shown in Figure 2 and Figure 3. In this case, only the existing WIFI equipment in the home environment is needed, and no additional professional equipment is needed. For example, without adding additional smoke detectors, the signal changes caused by fire can be analyzed to determine the current environment in the room. Is there a fire judgment and an alarm? The invention requires only one transmitting end and one receiving end. As shown in Figure 2, if there is a fire in the room, the fire will reflect the signal, and the CSI data received at the receiving end will change. The relevant features of the received CSI data can be used to determine whether the indoor is not There is a fire, and if there is, an alarm will be given.

In this example, the CSI data is used as an indicator. The CSI is short for Channel State Information, that is, channel state information, which can reflect the channel attribute of the communication link. The CSI data can indicate the weakening and scattering of the signal during the propagation process. The combined effects of etc. Based on the realization of the wireless propagation model in the current environment of the room, this example establishes the connection between CSI data and fire. In a stable indoor environment, CSI data is only affected by multipaths such as ceilings, floors and furniture. When the environment remains stable, its CSI data is also stable. If there is a fire in the room, the wireless signal will be reflected, which will affect the received CSI data. The CSI data is transmitted by using Orthogonal Frequency Division Multiplex (OFDM), and is decomposed into 30 subcarriers at the receiving end. By analyzing the number of links × subcarrier group data, it will be Room The situation is relatively accurate. That is, in the step S1, for each spatial stream, there are 30 subcarriers, which have different effects on the subcarrier values due to the frequency selective weakening.

Specifically, step S1 in this example includes the following sub-steps:

Step S11, collecting CSI data in a unit time period in the current environment, and obtaining a matrix of the number of links × the number of subcarriers;

Step S12, using a Butterworth low-pass filter (Butterworth algorithm) to perform low-pass filtering on the CSI data to remove the influence of Gaussian white noise;

Step S13, determining whether the mean square error is within the first threshold range by calculating the mean and the mean square error of each subcarrier in the unit time period, and if the first threshold value range is exceeded, determining whether the current environment has a person's activity or an unstable factor Interference, wait until the person leaves or the current environment returns to a stable state, and then jumps to step S2. The first threshold range is preferably greater than 5.76, and may also be customized and adjusted according to actual conditions (including different routers and different network cards).

In the step S13, the CSI data collected in the unit time period T (seconds) is a set of two-dimensional arrays and a matrix of size 90*(T*SAMPLES).

SAMPLES is the sampling point per second in the CSI data unit time period. First, the overall mean of the entire sequence over this period of time is derived.

And calculating the time series mean of the matrix

Where i is the sequence value of the sampling point, and the one-dimensional vector {average ₁ ,...average _i ,...average _T*SAMPLES } is the mean value in the unit time period. Calculation

Then, the variance sequence in the time period T corresponding to the time point t is obtained, and the mean square error is obtained.

In a stable environment, a wireless transmitter (AP) transmits a radio frequency signal, and a computer equipped with an Intel 5300 network card receives the CSI data as a wireless receiver. In the test experiment, the receiving end (the network card on the computer side) is equipped with 3 antennas, and the transmitting end (the AP end) has 1 antenna, forming three spatial links, and transmitting 30 subcarriers per day on the link, so we can receive one. 1 × 3 × 30 matrix. In order to remove the influence of white Gaussian noise, this example is adopted The Butterworth low-pass filter performs denoising on Gaussian white noise.

Step S2 in this example is used to calculate the probability of occurrence of data, and then calculate the angle of arrival, amplitude characteristics, and phase difference characteristics. Specifically, the step S2 includes the following sub-steps:

Step S21, extracting CSI data in a time period T0, and in this time period T0, the data of each subcarrier is first converged, that is, the value in the range of [X(i)-Θ, X(i)+Θ] Both are considered to be X(i), and X(i)<X(i+1), i∈Z, Θ is 0.01, and can also be customized and adjusted according to the actual situation. The smaller the Θ, the smaller the granularity, The more sensitive the environment, X(i) is the currently appearing data (or the currently occurring value); then the statistics on the number of occurrences of the data, and the combination of the data appearing in sequence [currently appearing data X(i), under The number of occurrences of an emerging data X(j)] is counted, and each different data X(i) and a different data combination are calculated [currently appearing data X(i), next occurrence of data X(j) )] The cumulative frequency that appears, by the cumulative frequency, respectively, the probability of occurrence of n different data X(i) and m different data combinations [currently occurring data X(i), the next occurrence of data X ( The probability of occurrence of j)], the state space probability and the joint probability are obtained; finally, the combination of interference data and interference data with probability of less than or equal to 1% is removed, that is, the probability is small The probability equal to 1% is set to 0; then, the probability is recalculated as the state space probability P={P _X(0) ,...,P _X(i) ,... } and joint probability Q={Q _Y(0) ,...,Q _Y(i) ,...}, and record it;

Step S22, using the MUSIC algorithm, the MUSIC algorithm is a method based on matrix feature space decomposition, and the ratio between the phase difference and the antenna spacing is calculated by the phase difference between the same subcarrier and the adjacent antenna, and the subcarrier is obtained. The angle of arrival, as the angle of arrival of the current environment in a steady state, and recorded;

Step S23, recording amplitude characteristics and phase difference characteristics of the current environment in a steady state;

Step S24, determining whether the mean square error is within the first threshold range by calculating the mean value and the mean square error of each subcarrier in the time period T0. If the first threshold value range is exceeded, determining whether the current environment has a person's activity or an unstable factor Interference, wait until the person leaves or the current environment returns to a stable state, and then jumps to step S3. The working principle of step S24 in this example is the same as that of step S13.

In the step S22, the calculation process of the arrival angle and the phase difference is as follows:

Known phase shift function between antennas

Where f is the frequency of the signal, c is the speed of light constant, θ _k is the angle of the arrival angle of the kth path, j is the imaginary number, and d is the antenna spacing (ie the spacing between the antennas). Known phase offset function between subcarriers

Where f _δ is the current subcarrier frequency and τ _k is the propagation time of the signal in the path.

Building vector

The data at this time is converted as follows.

A CSI smoothing matrix X is obtained, and the eigenvector ΧΧ ^{H of} the matrix is calculated, and a value close to 0 in the eigenvector ΧΧ ^H is determined as a noise vector E _N . By traversing all values of θ _k =-90, -89.9,...0,...89.9,90 and all of τ _k =0,10 ^-10 , 2*10 ^-10 ,...,10 ^-8 Value, calculation

The value in , when P _MU (θ _i , τ _j ) is much larger than the values of other P _MU (θ, τ), then θ _i is the angle of arrival, and Φ(θ _i ) is the phase, through the same subcarrier and phase The phase difference between adjacent antennas gets its phase difference.

In the step S23, the amplitude feature includes a distribution of amplitude values and a distribution of variance values of the size of the window in the sequence; the phase features include a distribution of phases.

Step S3 in this example continuously collects CSI data for the current environment, calculates the angle of arrival, and obtains the probability of occurrence of this condition P _codition1 ; by comparing the current state with the probability of each value and combination in the steady state state, This condition fire probability P _codition2 ; through the random judgment, the conditional fire probability P _codition3 is calculated; finally, the total fire occurrence probability P _fire =P _codition1 +P _codition2 +P _{codition3 is} calculated by calculation, thereby judging Whether the room is on fire.

In more detail, step S3 described in this example includes the following sub-steps:

Step S31, extracting CSI data in a unit time T1, where T1<<T0, by calculating each sub- The average and the mean square error of the carrier in the unit time T1, determining whether the mean square error is within the first threshold range, and if the first threshold range is exceeded, determining that the current environment has human activity or unsteady factor interference, waiting until the person leaves or After the current environment returns to a stable state, the process goes to step S32; the working principle of the step S31 is the same as that of the step S13;

Step S32, loading tolerance

Using the MUSIC algorithm, the phase difference between the antenna and the antenna is calculated by the phase difference between the adjacent subcarriers, and the arrival angle γ of the subcarrier is calculated.

in case

The probability of a fire in the current environment is

in case

Then the fire occurrence probability of the current environment is P _fire =P _codition1 =T _codition1 ; wherein γ ₀ is the average of the arrival angles in the steady state,

To tolerate the range of angular deviations (ie tolerance), T _codition1 is the weight by which the angle of arrival is calculated by calculating the angle of arrival, the tolerance

A preferred range of values is from 0.1 to 0.6, and a preferred range of the weight T _codition1 is from 0.1 to 0.4.

Step S33, extracting CSI data in the unit time T1, and in the unit time T1, the data of each subcarrier is first converged, that is, in the range of [X'(i)-Θ, X'(i)+Θ] The value is considered to be X'(i), and X'(i)<X'(i+1), i∈Z, Θ is 0.01, and can also be customized and adjusted according to the actual situation. The smaller, the more sensitive the environment, X'(i) is the currently appearing data (or the currently occurring value); then the number of occurrences of the data X'(i) is counted, and the data combinations appearing in sequence are [ The number of occurrences of the currently appearing data X'(i), the next occurrence of the data X'(j)] is counted, and each different data X'(i) and a different data combination are calculated. [Currently present data X'(i), the cumulative frequency of the next occurrence of the data X'(j)], and the cumulative probability of the occurrence probability of n' different data and the probability of occurrence of a different data combination of m' are calculated by the cumulative frequency. Obtain the state space probability and the joint probability; finally, the combination of the interference data and the interference data with the probability of less than or equal to 1% is removed, as the state of the current environment in the steady state is empty. The probability P'={P' _X(0) ,...,P' _X(i) ,...} and the joint probability Q'={Q' _Y(0) ,...,Q' _{Y( i)} , ...}; the working principle of the step S33 is the same as the working principle of the step S21;

Step S34, using a chi-square test method, calculating

with

N is the number of elements different from X(i) in P and P', M is the number of elements different from Y(i) in Q and Q', and X(s) is the smallest element in P and P' The element, X(e) is the largest element of the elements in P and P', Y(s) is the element with the smallest element in Q and Q', and Y(e) is the element with the largest element in Q and Q'; The current confidence interval α _{1 of the} chi-square test is compared by 查₁ ² and χ ² (N-1) and χ ₂ ² and χ ² (M-1) by looking up the table (Table 1 card-side distribution critical table). Size, if and only if χ ₁ ² >χ ² (N-1) and χ ₂ ² >χ ² (M-1) is established, proceed to step S35, otherwise,

Calculating _{P fire} = P codition1 + P _codition2 is greater than Δ _ALARM, is greater than the alarm, otherwise, returns to step S31 to continue to monitor; _wherein, T codition2 is a weight comparison determination fire probability by the probability that the environment and the steady state current _weight, T codition3 to determine the probability of a fire by weight of randomness, Δ _aLARM to whether the alarm threshold value; wherein the weight _codition2 T is preferably in the range of 0.1 to 0.5; the weight is preferably in the range of 0.1 to 0.5 _T codition3 The preferred range of the threshold value _ΔALARM is from 0.6 to 0.99; of course, these ranges of values are preferred and can be adjusted according to actual conditions.

Table 1 card square distribution critical table

Step S35, checking the randomness, using the run-length test method, dividing the elements in the two difference sets P-P' and Q-Q' into two parts according to the frequency average, so that the two parts of the frequency are the same and the part is the same. The element is 1, and the other part is 0, which is substituted into the CSI data sequence, and the sequence containing only 0 and 1 is extracted, and the same value appears consecutively in the sequence (ie, 0 or 1 respectively appear one or consecutively multiple) The number of times, recorded as the number of runs r, the number of occurrences of 1 is recorded as n ₁ , the number of occurrences of 0 is recorded as n ₂ , and the mean value of the sample distribution is calculated.

And sampling distribution variance

Calculate run statistics

Use the run statistic to check the table (Table 2 standard positive distribution table). In this table 2, the integer digit of the behavior Z and the first digit of the decimal point are listed as the second decimal place of Z, and then the current probability p is obtained by looking up the table. Loading the confidence interval α ₂ used in the current run test to determine whether p<1-α ₂ is satisfied. If it is not satisfied, it is considered that the randomness is not true, and the process returns to step S31. If it is satisfied, the randomness is considered to be established, and P _fire = P is calculated. _codition1 + P _codition2 + P _codition3 is greater than Δ _aLARM, is greater than the alarm, otherwise, returns to step S31 to continue to monitor; wherein the confidence interval α ₂ preferably ranges from 0.001 to 0.2.

Table 2 standard positive distribution table

The step S31 to the step S35 are the necessary sub-steps of the step S3; preferably, in addition to the step S31 to the step S35, the step S3 may further include the step S36 to implement the correction of the parameter.

In step S36, a fire alarm signal is issued. If the alarm is not turned off in time, a message is sent to the owner. If the owner determines the alarm as a false alarm, the parameter is traversed by the feasible value of the previous parameter and the current CSI data. T _codition1 , T _codition2 , T _codition3 , T _{codition4 ,} and Δ _ALARM are corrected.

Substituting all the data of all the parameters in the P _fire when all the fires are issued into the original formula P _fire =P _codition1 +P _codition2 +P _codition3 , let the vector P _{fire_wrong} <0.6 and the vector P _{fire_right} >0.6, calculate all T _codition1 , _T codition2, T codition3, T codition4 Δ _ALARM and feasible solution is to find out where all the current parameters and change the minimum feasible solution, reset parameter; _wherein, P fire_wrong of false fire _vector, P fire_right correct trigger fire vector.

Preferably, the number of transmitting ends in the example is 1, and the number of the receiving ends is 3 or more.

In the case of stable environment, the subcarrier signal changes fluctuate within a certain range, and the probability distribution is stable. Since the signal is affected by the physical properties of the flame, the generation of the fire will cause the stable environment to be broken, and a new state or a new distribution will appear, which is a judgment condition for the occurrence of the fire.

In the flame, part of the air is plasma, the part of the air will reflect the signal, and the plasma part of the flame will flow by the flow of the air, when the plasma part flows, the plasma The curvature will change or because the flow causes the plasma to be out of the signal propagation path, so the signal on other paths will increase or decrease, and the superposition of the multipath will change, which causes the arrival angle of different subcarriers to occur. Change, can be analyzed by MUSIC algorithm, there is a significant change in the angle of arrival, which is a judgment condition for the occurrence of fire.

Because the plasma portion of the flame is generated by the flow of the airflow, the fluctuation is random, which is also a condition for the occurrence of the fire.

Calculating the magnitude of the subcarrier amplitude variation can be used to distinguish between the effects of humans on the signal and the effects of fire on the effects. Because the human occlusion area is relatively large, and the human activity displacement/signal wavelength is relatively large, the influence amplitude of the signal on the signal changes sharply and rapidly, while the transmission and reflection of the fire are relatively small, so the person The activity is not the same as the effect of fire on the signal. This is also a condition for judging whether there are people or other unsteady factors in the current environment.

Specifically, as shown in FIG. 4, the data processing process of the indoor fire detection and alarm method based on wireless signal transmission in this example is mainly divided into three parts: denoising, extracting features, and probability estimation.

The present invention also provides a system for indoor fire detection and alarm based on wireless signal transmission, and the system for indoor fire detection and alarm based on wireless signal transmission applies the above-mentioned indoor fire detection and alarm based on wireless signal transmission. Methods, including:

a CSI data acquisition module, configured to receive a radio frequency signal from the transmitting end, calculate CSI data, and remove a Gaussian white noise by using a Butterworth low-pass filter;

The data processing module separately performs statistical probability on the data, and calculates its angle of arrival, chi-square test and run-length test;

The fire condition judging module calculates the possibility of the above data, and determines whether the current environment has a fire by whether it is in the second threshold range;

An alarm module, configured to issue a fire alarm signal and notify the owner when a fire occurs in the current environment;

The feedback correction module is configured to correct the parameters T _codition1 , T _codition2 , T _codition3 , T _{codition4 ,} and Δ _ALARM when the alarm is determined to be a false alarm.

As shown in FIG. 5, the system for indoor fire detection and alarm based on wireless signal transmission in this example is mainly divided into four parts: environment awareness, data processing, alarm and feedback. More specifically, as shown in FIG. 6, the process for realizing indoor fire detection and alarm includes:

1. The transmitting end (wireless AP) sends a wireless signal, and the receiving end (a computer with a wireless network card) receives the wireless signal and calculates the CSI data;

2. Each link has 30 subcarriers that can be used to characterize channel state information;

3. Perform low-pass filtering on the collected CSI signals by using the Butterworth algorithm to perform denoising;

4. The mean square error judgment algorithm calculates the mean square error;

5. Calculate the angle of arrival using the MUSIC algorithm;

6. Calculate the probability of the value of each subcarrier;

7. Use the chi-square test to check the coincidence degree of each sub-carrier;

8. Use the method of run test to test randomness;

9. Calculate whether the probability of occurrence of the fire exceeds the threshold of the alarm;

10. If there is a fire, issue an alarm. If the alarm is not closed in time, send a text message to the owner.

The second threshold range described in this example is preferably 0.6, and can also be customized and adjusted according to actual conditions.

More specifically, the CSI data acquisition module in this example includes:

a sensing unit, configured to initialize channel state data, and collect CSI data, where each CSI data is a matrix of link number×subcarrier number;

The filtering unit uses the Butterworth algorithm to remove the influence of Gaussian white noise on the channel state data.

The data processing module in this example is used to obtain an average value of CSI data of 30 consecutive subcarriers at the same time point for each spatial stream, and the average value is used as channel state data;

a data feature calculation unit, configured to extract features for CSI data, the extracted features include performing mean square error determination, statistical probability, calculating an angle of arrival, calculating a chi-square test, and calculating a run-length test on the CSI data;

The first abnormal output unit performs a calculation on the probability of interference of the human activity on the data, and determines whether it is within a limited first threshold range, thereby determining whether there is interference of the human activity or the unsteady factor in the room;

The second abnormality output unit calculates the fire possibility of the data, and determines whether there is a fire in the room by whether it is within a limited second threshold range.

The data feature calculation unit described in this example is used to extract features for CSI data, and is divided into five steps:

Step 1: In the CSI data, calculate the mean value and the mean square error of each subcarrier in a unit time period, determine whether the mean square error is within the first threshold range, and if the first threshold value range is exceeded, determine the current environment existing person activity or other If the non-stabilizing factor interferes, the waiting person leaves or the environment returns to a stable state, and then the next feature extraction is performed;

Step 2: Calculate the value of each subcarrier for probability and statistics, and calculate the transition probability between the values;

Step 3: Using the MUSIC algorithm, calculate the angle of arrival of the multipath, and determine whether the signal changes on the original path. If there is a change, there may be a possibility of fire;

Step 4: Using the chi-square test to check the degree of coincidence of each sub-carrier, if most of them are not coincident, there may be a possibility of fire;

Step 5: Use the method of run test to test the randomness.

The fire condition judging module of the present example includes a fire condition judging unit that calculates the possibility of the above data, and determines whether there is a fire in the room by whether it is within a limited second threshold range.

The alarm module of this example includes an alarm unit, and after issuing a fire alarm signal and notifying the homeowner, if the alarm is not responded or eliminated in time, a help signal is sent to the police.

In the feedback correction module of this example, all feasible solutions of related parameters are saved, the values of the relevant parameters are adjusted by the current CSI data, and the combination of feasible solutions is reduced to increase the accuracy, and the related parameters include T _codition1 , T _codition2 , T _codition3 , P _codition3 and Δ _ALARM , the relevant parameters referred to in this example are referred to as parameters.

The above is a further detailed description of the present invention in connection with the specific preferred embodiments, and the specific embodiments of the present invention are not limited to the description. It will be apparent to those skilled in the art that the present invention may be made without departing from the spirit and scope of the invention.

Claims (10)

1. A method for indoor fire detection and alarm based on wireless signal transmission, characterized in that it comprises the following steps:

   In step S1, the receiving end receives the radio frequency signal from the transmitting end, obtains the CSI data, uses the Butterworth low-pass filter to remove the Gaussian white noise, and determines whether there is human or non-stabilizing factor interference in the current environment by using the variance method. Then, it is determined that the current environment is in an unstable state, waiting for the current environment to enter a steady state, if otherwise, directly jump to step S2;

   Step S2, respectively performing statistical probability on the current environment as a steady state data, and calculating an angle of arrival, an amplitude characteristic, and a phase difference characteristic;

   Step S3, continuously collecting the CSI data of the current environment, calculating the angle of arrival, and _{obtaining the} probability of occurrence of the fire in the current environment P _codition1 , and comparing the probability of the current environment with the steady state to obtain the fire probability P _{codition2 of the} current environment. The current environmental fire probability P _{codition3 is} calculated by randomness judgment. Finally, the total fire occurrence probability P _fire =P _codition1 +P _codition2 +P _{codition3 is} calculated by calculation to determine whether the room is on fire.
2. The method for detecting and alarming indoor fire conditions based on wireless signal transmission according to claim 1, wherein said step S1 comprises the following substeps:

   Step S11, collecting CSI data in a unit time period in the current environment, and obtaining a matrix of the number of links × the number of subcarriers;

   Step S12, performing low-pass filtering processing on the CSI data by using a Butterworth low-pass filter to remove the influence of the white Gaussian noise;

   Step S13, determining whether the mean square error is within the first threshold range by calculating the mean and the mean square error of each subcarrier in the unit time period, and if the first threshold value range is exceeded, determining whether the current environment has a person's activity or an unstable factor Interference, wait until the person leaves or the current environment returns to a stable state, and then jumps to step S2.
3. The method for detecting and alarming indoor fire based on wireless signal transmission according to claim 1, wherein said step S2 comprises the following substeps:

   Step S21, extracting CSI data in a time period T0. In this time period T0, the data of each subcarrier is first converged, then the number of occurrences of the data is counted, and the number of times the data combination occurs in sequence Statistics are performed to calculate the cumulative frequency of each data and data combination. The probability of occurrence of n different data and the probability of occurrence of m different data combinations are obtained by the cumulative frequency, and the state space probability and joint probability are obtained. Finally, the combination of interference data and interference data with a probability less than or equal to 1% is removed as the state space probability P={P _X(0) ,...,P _{X(i) of the} current environment in a steady state. ...} and joint probability Q={Q _Y(0) ,...,Q _Y(i) ,...}, and record;

   Step S22, using the MUSIC algorithm, calculating the ratio between the phase difference and the antenna spacing by using the phase difference between the same subcarrier and the adjacent antenna, and obtaining the arrival angle of the subcarrier, thereby using the current environment as the steady state. Arrival angle and record;

   Step S23, recording amplitude characteristics and phase difference characteristics of the current environment in a steady state;

   Step S24, determining whether the mean square error is within the first threshold range by calculating the mean value and the mean square error of each subcarrier in the time period T0. If the first threshold value range is exceeded, determining whether the current environment has a person's activity or an unstable factor Interference, wait until the person leaves or the current environment returns to a stable state, and then jumps to step S3.
4. The method of indoor fire detection and alarm based on wireless signal transmission according to claim 3, wherein the step S3 comprises the following sub-steps:

   Step S31, extracting CSI data in a unit time T1, wherein T1<<T0, determining whether the mean square error is within the first threshold range by calculating the mean and the mean square error of each subcarrier in the unit time T1, if The first threshold range, it is determined that the current environment exists human activity or non-stability factor interference, wait until the person leaves or the current environment returns to a stable state, then jumps to step S32;

   Step S32, loading tolerance

   

   Using the MUSIC algorithm, the phase difference between the antenna and the antenna is calculated by the phase difference between the adjacent subcarriers, and the arrival angle γ of the subcarrier is calculated.

   

   in case

   

   The probability of a fire in the current environment is

   

   in case

   

   Then the fire occurrence probability of the current environment is P _fire =P _codition1 =T _codition1 ; wherein γ ₀ is the average of the arrival angles in the steady state,

   

   To tolerate the range of angular deviations, T _codition1 is the weight that determines the probability of occurrence of a fire by calculating the angle of arrival;

   Step S33, extracting CSI data in unit time T1, in this unit time T1, first converging data of each subcarrier, then counting the number of occurrences of the data, and performing the number of occurrences of the data combinations appearing in sequence Statistics, calculate the cumulative frequency of each data and data combination, and obtain the probability of occurrence of n' different data and the probability of occurrence of data combination with different m' by cumulative frequency, and calculate the state space probability and joint probability. Finally, the combination of interference data and interference data with a probability less than or equal to 1% is removed as the state space probability P'={P' _X(0) ,...,P' _{X of the} current environment in a steady state. _(i) ,...} and joint probability Q'={Q' _Y(0) ,...,Q' _Y(i) ,...};

   Step S34, using a chi-square test method, calculating

   

   with

   

   N is the number of elements different from X(i) in P and P', M is the number of elements different from Y(i) in Q and Q', and X(s) is the smallest element in P and P' The element, X(e) is the largest element of the elements in P and P', Y(s) is the element with the smallest element in Q and Q', and Y(e) is the element with the largest element in Q and Q'; The current chi-square test uses the confidence interval α _{1 to} compare the sizes of χ ₁ ² and χ ² (N-1) and χ ₂ ² and χ ² (M-1) by looking up the table, if and only if χ ₁ ² >χ ² (N-1) and χ ₂ ² >χ ² (M-1) is established, proceed to step S35, otherwise,

   

   Calculating _{P fire} = P codition1 + P _codition2 is greater than Δ _ALARM, is greater than the alarm, otherwise, returns to step S31 to continue to monitor; _wherein, T codition2 is a weight comparison determination fire probability by the probability that the environment and the steady state current _weight, T codition3 In order to determine the weight of the fire probability by randomness, _ΔALARM is the threshold of whether or not to alarm;

   Step S35, checking the randomness, using the run-length test method, dividing the elements in the two difference sets P-P' and Q-Q' into two parts according to the frequency average, so that the two parts of the frequency are the same and the part is the same. The element is 1 and the other part is 0, which is substituted into the CSI data sequence, and the sequence containing only 0 and 1 is extracted. The number of consecutive occurrences of the same value in the sequence is recorded as the run number r, and the number of 1 will appear. Recorded as n ₁ , the number of occurrences of 0 is recorded as n ₂ , and the mean value of the sampling distribution is calculated.

   

   And sampling distribution variance

   

   Calculate run statistics

   

   The confidence interval α ₂ used for loading the current run test is used to perform a table lookup using the run statistic to obtain the current probability p, and it is judged whether or not p<1-α ₂ is satisfied. If not, the randomness is not satisfied, and the process returns to step S31, if it is satisfied. , it is established that random, then _P codition3 = _T codition4, calculate _{P fire = P codition1 + P codition2} + P codition3 is greater than Δ _aLARM, is greater than the alarm, otherwise, returns to step S31 to continue to monitor; _wherein, T codition4 through The probability of the fire probability is judged by comparing the probability of the current environment with the steady state;

   In step S36, a fire alarm signal is issued. If the alarm is not turned off in time, a message is sent to the owner. If the owner determines the alarm as a false alarm, the parameter is traversed by the feasible value of the previous parameter and the current CSI data. T _codition1 , T _codition2 , T _codition3 , T _{codition4 ,} and Δ _ALARM are corrected.
5. The method for detecting and alarming indoor fire based on wireless signal transmission according to any one of claims 1 to 4, wherein the number of the transmitting ends is 1, and the number of the receiving ends is three or more. .
6. A system for indoor fire detection and alarm based on wireless signal transmission, characterized in that it comprises:

   a CSI data acquisition module, configured to receive a radio frequency signal from the transmitting end, calculate CSI data, and remove a Gaussian white noise by using a Butterworth low-pass filter;

   The data processing module separately performs statistical probability on the data, and calculates its angle of arrival, chi-square test and run-length test;

   The fire condition judging module calculates the possibility of the above data, and determines whether the current environment exists by whether it is in the second threshold range. Have a fire;

   An alarm module, configured to issue a fire alarm signal and notify the owner when a fire occurs in the current environment;

   The feedback correction module is used to correct the parameters when the alarm is determined to be a false alarm.
7. The system for detecting and alarming indoor fire detection based on wireless signal transmission according to claim 6, wherein the CSI data acquisition module comprises:

   a sensing unit, configured to initialize channel state data, to obtain a matrix of link number×subcarrier number;

   The filtering unit uses the Butterworth algorithm to remove the influence of Gaussian white noise on the channel state data.
8. The system for detecting and alarming indoor fire based on wireless signal transmission according to claim 6, wherein the data processing module comprises:

   a data feature calculation unit, configured to extract features for CSI data, the extracted features include performing mean square error determination, statistical probability, calculating an angle of arrival, calculating a chi-square test, and calculating a run-length test on the CSI data;

   The first abnormal output unit performs a calculation on the probability of interference of the human activity on the data, and determines whether it is within a limited first threshold range, thereby determining whether there is interference of the human activity or the unsteady factor in the room;

   The second abnormality output unit calculates the fire possibility of the data, and determines whether there is a fire in the room by whether it is within a limited second threshold range.
9. The system for detecting and alarming indoor fire based on wireless signal transmission according to claim 6, wherein in the alarm module, after the fire alarm signal is issued and the owner is notified, if the alarm is not timely responded or eliminated, In turn, a help signal will be sent to the police.
10. The system for detecting and detecting indoor fire detection based on wireless signal transmission according to claim 6, wherein the feedback correction module stores a feasible solution of all relevant parameters, and the current CSI data is used to correlate the relevant parameters. The values are adjusted and the combination of their feasible solutions is reduced to increase the accuracy, including T _codition1 , T _codition2 , T _codition3 , T _{codition4 ,} and Δ _ALARM .

Images