Indoor personnel passive sensing and intrusion detection method based on CSI amplitude stationarity analysis

Abstract

The application discloses an indoor personnel passive sensing and intrusion detection method based on channel state information (CSI) amplitude stationarity analysis. The method comprises the following steps: collecting samples under indoor personnel and no personnel conditions respectively; extracting CSI amplitude matrix from the samples and carrying out denoising and abnormal value processing; calculating global stability and local stability of the CSI amplitude matrix; and realizing passive sensing and intrusion detection of indoor personnel according to a double-threshold joint determination algorithm. The application uses Wi-Fi signals as sensing signals, has low cost and simple system construction; the application uses CSI as a feature, has the advantages of fine granularity and good time stability, the technology has high concealment, is not limited by the environment, and greatly improves sensing space and applicability; the application adopts wavelet denoising and Hanning filter processing methods to process data, has the characteristics of multi-resolution analysis, and improves the anti-interference property of the system; and the double-threshold joint determination algorithm realizes passive sensing and intrusion detection of indoor personnel in all states.

Description

Technical Field

[0001] This invention relates to a passive sensing and intrusion detection method based on Wi-Fi signals and utilizing the CSI amplitude stability to determine the presence of people indoors. Background Technology

[0002] Currently, with the rapid development of mobile communication technology and the sensor network industry, intelligent sensing technology has become an indispensable technology in social life. Traditional indoor sensing and intrusion detection methods mainly include video recording, infrared sensing, wearable devices, RFID, ZigBee, etc. However, some of these sensing methods pose risks to user privacy in real life, and their application is easily limited by the environment. These methods may require special hardware, be easily affected by light and smoke, or have problems such as complex wiring and low recognition rate, making it difficult for them to be widely used.

[0003] Research has found that in indoor settings, radio waves undergo various refractions, reflections, and scattering due to people or objects during propagation, resulting in multipath superposition signals at the receiver. These signals are affected by the indoor environment and therefore carry information reflecting environmental characteristics. Compared to traditional indoor sensing methods, wireless signal sensing technology is more resistant to environmental changes, has higher concealment, requires no devices worn by the target, and can bypass complex indoor obstacles, greatly improving the sensing space and applicability. It is worth noting that indoor WLAN devices have become very common, and considering factors such as signal coverage area, concealment, and deployment costs, Wi-Fi signals have become the preferred signal for indoor sensing technology.

[0004] In early research on indoor people detection and intrusion detection based on WLAN, RSSI was often used as an indicator parameter of WLAN wireless signals due to its easy availability. Some foreign scholars proposed a system based on the mean and variance of RSSI to detect intrusion behavior between the transmitter and receiver; however, tests in real-world applications revealed a performance degradation in this system. Subsequently, the RASID system further improved the performance of intrusion detection systems by analyzing the characteristics of RSSI and employing a non-parametric technique. Although RSSI-based research has achieved some results, overall, RSSI-based solutions still suffer from drawbacks such as low detection rates and high false alarm rates.

[0005] In recent years, Channel State Information (CSI) has been readily available using Intel 5300 series or other wireless network cards under the 802.11n protocol. Compared to RSSI, CSI offers advantages such as resistance to frequency-selective fading and more detailed characterization of multipath effects under IEEE 802.11n channels, making it more suitable as an indicator parameter for WLAN wireless signals. Therefore, most recent research on indoor occupant sensing and intrusion detection is based on CSI. Leveraging the fine-grained and high-sensitivity advantages of CSI, researchers can utilize it to perform many subtle sensing tasks, such as sleep apnea monitoring, indoor positioning, and activity recognition.

[0006] Previously, researchers both domestically and internationally proposed using phase information from CSI to extract feature values ​​and compare them with a standard stable set to determine intrusion behavior. However, this method only targets intrusion behavior and is difficult to detect indoor personnel in slow-moving or stationary states. Furthermore, intrusion detection methods using temporal phase features require secondary processing of the collected CSI data to extract phase information. In comparison, amplitude-based methods are more lightweight.

[0007] Addressing the shortcomings of traditional indoor personnel sensing methods and the rapid development of wireless LANs, this invention proposes using existing commercial Wi-Fi devices to process the channel state information of Wi-Fi signals, thereby achieving seamless and passive indoor personnel sensing and intrusion detection. This invention eliminates the need for equipment modification, deployment of additional sensors, limitations on line-of-sight environments, or continuous movement of the detected personnel, enabling passive indoor personnel sensing and intrusion detection in various typical scenarios. Summary of the Invention

[0008] In view of the shortcomings of the above methods, this invention proposes an indoor passive sensing and intrusion detection method based on CSI amplitude stationarity analysis. It overcomes the problems of traditional methods, such as the need to modify hardware equipment, susceptibility to light and smoke, complex wiring, and the requirement for continuous movement of the detected personnel. At the same time, it solves the drawbacks of low recognition rate and high false alarm rate when using RSSI as an indicator parameter, and achieves high-precision indoor passive sensing and intrusion detection.

[0009] The technical solution for implementing the present invention is as follows:

[0010] This method receives Wi-Fi signal channel status information under different indoor scenarios and determines whether intrusion has occurred or whether anyone is present in the current indoor environment based on CSI stationarity analysis. The specific steps are as follows:

[0011] (1) Collect Wi-Fi channel status information data in indoor environments with and without people to form a sample library;

[0012] (2) Extract the CSI amplitude matrix from the original sampled data and perform noise reduction and outlier processing;

[0013] (3) Calculate the global and local stability of the CSI amplitude matrix over a period of time;

[0014] (4) The global stability GS in the current state is determined by a dual-threshold joint determination algorithm. NOW With local stability LS NOW The results are compared with the set thresholds l1 and l2 to determine whether there are people present in the room or whether there has been an intrusion.

[0015] Furthermore, this invention requires only one router to transmit a Wi-Fi signal. The Wi-Fi signal is transmitted by the deployed router AP under the 802.11n protocol, and the Wi-Fi signals collected in the same indoor environment all come from the same router.

[0016] Furthermore, the Wi-Fi channel status information collected in step (1) of this invention is data over a period of time, not data at a specific moment. Also, when collecting data indoors with people present, there is no limit to the number of people indoors; the number of people indoors can be greater than one.

[0017] Furthermore, the sample library described in this invention includes channel state information data collected in different locations, at different times, under line-of-sight and non-line-of-sight conditions, and under different personnel movement states.

[0018] Furthermore, the Channel State Information (CSI) described in this invention is a three-dimensional complex matrix of size M×N×K at each time step. The first dimension M represents the number of transmit antennas, the second dimension N represents the number of receive antennas, and the third dimension K represents the number of subcarriers. Each element in the matrix is ​​a complex number. Therefore, the CSI received by the j-th receive antenna at time t from the i-th transmit antenna can be expressed as:

[0019]

[0020] in, It means that the receiving antenna j receives data from the k-th subcarrier of the i-th transmitting antenna at time t, each All can be represented as:

[0021]

[0022] The CSI amplitude vector received by antenna j from the i-th transmitting antenna at time t can be extracted and expressed as:

[0023]

[0024] The CSI amplitude vectors at different times are combined to form a CSI amplitude matrix over a time interval of m for analysis, observing its changes over time. The CSI amplitude matrix H over the time interval is... i，j It can be represented as:

[0025]

[0026] Furthermore, in step (2) of this invention, wavelet threshold denoising is used to suppress signal noise. This method is simple to implement, has good denoising effect, and is very fast in calculation. Compared with other filtering methods, it has the characteristics of multi-resolution analysis, which can focus on any detail of the signal to perform multi-resolution time-frequency domain analysis, and can well preserve the detailed information of the signal.

[0027] Furthermore, the amplitude of the CSI data collected in practice may exhibit irregular jumps. This phenomenon could be caused by factors such as indoor noise, unstable signals from the AP device itself, and network card hardware issues. These abnormal data jumps will inevitably affect subsequent amplitude stability analysis, thus impacting the sensing and detection results. Therefore, to ensure the accuracy of indoor personnel sensing and intrusion detection, further work is needed on outliers after noise reduction. This invention uses a Hamper filter to detect and replace outliers in the CSI amplitude time-series data.

[0028] Furthermore, step (3) of this invention proposes using two indicators, global stability and local stability, to perceive the current indoor environmental state. For scenarios where people are moving around indoors, this will inevitably cause a certain degree of disturbance to the CSI data within a certain time period. In this case, the average standard deviation of the amplitude vectors of N subcarriers within a time period m is calculated and called global stability, thus measuring the overall fluctuation of the sample. For scenarios where people are moving slowly or stationary indoors, the overall fluctuation of the CSI data within the current time period is not obvious, and the indication effect of global stability GS is poor. In this case, introducing the concept of local stability LS can provide a fine-grained description of the current environment and accurately perceive indoor people in a static state.

[0029] Furthermore, the dual-threshold joint judgment algorithm in step (4) of the present invention compares the two indicators reflecting the environmental state, global stability GS and local stability LS, with the set thresholds, combining information from both coarse and fine aspects to obtain the judgment result of indoor personnel perception.

[0030] Furthermore, if the current state indicates that someone is present and the room was empty at the previous moment, it is considered that there has been an intrusion. Intrusion detection of people inside the room is implemented based on this rule.

[0031] Compared with existing technologies, the indoor passive occupant sensing and intrusion detection method proposed in this invention has the following advantages:

[0032] (1) The sensing signal used in this invention is emitted by a widely available indoor Wi-Fi router. The indoor passive sensing and intrusion detection system can be built using only a home router and a computer with an Intel 5300 network card. The system requires no additional equipment purchase, resulting in extremely low cost; no further hardware modifications are required, making the system easy to set up.

[0033] (2) This invention utilizes the channel state information (CSI) of Wi-Fi signals for indoor personnel perception and intrusion detection. Compared with RSSI, it has the characteristics of fine granularity and high resolution. At the same time, CSI has the advantage of good time stability, which solves the drawbacks of low recognition rate and high false alarm rate when using RSSI as an indicator parameter, and realizes high-precision indoor passive personnel perception and intrusion detection.

[0034] (3) The indoor passive sensing technology described in this invention has strong concealment and does not require the target to wear a device, but can complete the detection and sensing of the target's intrusion behavior. At the same time, the method is not affected by environmental factors such as light and smoke, can bypass indoor obstacles, and can continue to work even under non-line-of-sight conditions, greatly improving the sensing space and applicability.

[0035] (4) The present invention uses wavelet threshold denoising and Hamper filtering to process the data, which is simple to implement and has a good denoising effect. It has the characteristics of multi-resolution analysis, which can focus on any detail of the signal to perform multi-resolution time-frequency domain analysis. It can preserve the detailed information of the signal well. At the same time, the Hamper filtering method is used to detect and replace abnormal data in the CSI amplitude time series data, which improves the anti-interference ability of the system.

[0036] (5) This invention combines global stability information GS with local stability information LS. The two types of information can accurately perceive different states of the target and complement each other. By using these two types of information together, we can achieve passive perception of the full state of indoor personnel and effectively reduce the missed detection rate of intrusion behavior in static or slow-moving states. Attached Figure Description

[0037] Figure 1 This is a flowchart of the indoor passive occupant perception and intrusion detection method based on CSI amplitude stationarity analysis of the present invention.

[0038] Figure 2 This is a diagram of the CSI data structure in the wireless transmission system used in this invention;

[0039] Figure 3 A graph showing the difference in CSI amplitude distribution when there are people in or out of the room;

[0040] Figure 4 This is a flowchart of the dual threshold determination algorithm. Detailed Implementation

[0041] The method described in this invention will be explained in detail with reference to the accompanying drawings and embodiments.

[0042] like Figure 1 As shown, the indoor passive occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis is as follows:

[0043] (1) Collect Wi-Fi channel status information data in indoor situations with and without people to form a sample library.

[0044] In indoor scenarios where routers are deployed, the distribution of wireless links within the space is stable or fluctuates slightly under fixed conditions. When people are within the detection area, they will inevitably cause some disturbance to the wireless links. By analyzing the link fluctuations over a period of time, it is possible to detect people and intrusion behavior within the indoor space.

[0045] This invention is based on an Ubuntu 14.04 LTS system equipped with an Intel 5300 network card and a router with three antennas to realize a 3-transmit and 3-receive wireless transmission system. The computer is set to AP mode to receive data, and the acquisition rate is set to a fixed value. Data corresponding to 30 subcarrier frequencies of the Wi-Fi signal transmitted by the router are collected. Therefore, the CSI data collected at each moment is a 3*3*30 matrix H.

[0046] To address complex indoor scenarios, Channel State Information (CSI) of indoor Wi-Fi signals is collected and stored on a computer under various conditions, including when the room is empty, with one person present, and with multiple people present. Each collection session lasts for milliseconds. Furthermore, considering the characteristics of non-cooperative intrusion actions being small in scale and not fixed in location, this invention also collects CSI data for indoor personnel in both moving and stationary states, and for the transceiver in both line-of-sight and non-line-of-sight states, forming an original reference sample library and a test sample library.

[0047] (2) Extract the CSI amplitude matrix from the original sample data and perform noise reduction and outlier processing.

[0048] For each sample in step (1), the CSI data packet collected at each time step is a 3*3*30 three-dimensional data, such as Figure 2 As shown, each H(i, j, k, t) is a complex number, equivalent to the CSI sample value at time t on the k-th subcarrier between the i-th antenna at the transmitting end and the j-th antenna at the receiving end. Therefore, the CSI received by the j-th receiving antenna at time t from the 1st transmitting antenna can be expressed as:

[0049]

[0050] in, The receiving antenna j receives data from the i-th subcarrier of the first transmitting antenna at time t. All can be represented as:

[0051]

[0052] Subsequently, the CSI amplitude received by antenna j from the first transmitting antenna at time t is extracted to form an amplitude vector, which can be represented as:

[0053]

[0054] Finally, the CSI amplitude vectors at different times are put together to obtain the CSI amplitude matrix over time m.

[0055] Since the acquired CSI matrix contains noise vectors, noise interference needs to be eliminated before subsequent processing. This invention employs wavelet threshold denoising to filter out noise. While preserving the low-frequency signal, wavelets are selected to determine the threshold parameter during the denoising process. Based on the threshold parameter selected in the previous step, the original signal is decomposed into w layers of wavelets using a wavelet function, thus filtering out background noise.

[0056] In addition, in practice, the channel state information may fluctuate irregularly due to factors such as unstable signal of the AP device itself and noise interference. This invention uses Hamper filtering to fill out outliers other than 3σ with the median value to complete the processing of outliers.

[0057] (3) Calculate the global stability and local stability of the CSI amplitude matrix over a period of time.

[0058] Based on the extraction, denoising and outlier processing of the CSI amplitude matrix in step (2), this invention takes two dimensions to deeply observe the distribution characteristics of the CSI amplitude matrix in the corresponding indoor scene during this period, and proposes to use two indicators, global stability (GS) and local stability (LS), to jointly reflect the current indoor environmental state.

[0059] In scenarios involving indoor movement, the activity of people will inevitably cause significant disturbances to the CSI data over a certain period of time. The distribution of CSI amplitudes in indoor environments with and without people is as follows: Figure 3 As shown. The average standard deviation of N subcarriers over a given time period represents the global stability. The global stability GS is obtained by calculating the average standard deviation of the CSI amplitude vector received by antenna j from the first transmitting antenna within time m, across 30 subcarriers, and can be expressed as:

[0060]

[0061] In scenarios where people are moving slowly or stationary indoors, the overall fluctuation of CSI data within the current time period is not significant, and the indication effect of using only global stability (GS) is no longer reliable. However, the chest rise and fall caused by human breathing and subtle movements can change the propagation path of indoor wireless signals, thereby causing fluctuations in the amplitude of CSI signals. In this case, introducing the concept of local stability (LS) can provide a fine-grained description of the current environment and accurately perceive indoor people in static conditions.

[0062] Steady-state detection is performed on 30 subcarriers within a duration of m seconds using a sliding window method of length W. The sample data of the k-th subcarrier is divided into a stable duration set Sd. k With fluctuation duration set Fd k If the window data CSIdata satisfies std(CSIdata) < r × mean(CSIdata), then this set of data is determined to be steady-state data and is incorporated into the steady-state duration set Sd. k In the middle, otherwise it is incorporated into the fluctuation duration set Fd k The window is moved sequentially forward until the steady-state status of all time-series data has been determined. The final local stability LS for the current time period can be expressed as:

[0063]

[0064] (4) By using the dual threshold joint judgment algorithm, the global stability and local stability under the current state are compared with the set thresholds to determine whether there are people in the room.

[0065] The dual-threshold joint judgment algorithm combines global stability (GS) and local stability (LS) to overcome the high false negative rate of single indicators in both dynamic and static scenarios. The algorithm process is as follows: Figure 4 As shown.

[0066] Perform steps (2) and (3) on the reference sample data collected in step (1) to obtain GS and LS datasets corresponding to indoor manned and unmanned states. Divide them into two classes and draw scatter plots. Determine the standard thresholds l1 and l2 through the clustering results.

[0067] The same steps (2) and (3) are performed on the CSI data collected in the current state to obtain the GS data for a certain period of time in the current state. NOW With LS NOW The index is compared with standard thresholds l1 and l2. If GS NOW Less than l1 and LS NOWIf the value is greater than l2, the current state is determined to be unoccupied; otherwise, the current state is determined to be occupied. Furthermore, if the previous state was determined to be unoccupied and the current state is determined to be occupied, an intrusion is considered to have occurred.

[0068] The above is merely a further description of the present invention and is not intended to limit the implementation and application of this patent. All equivalent implementations of the present invention should be included within the scope of the claims of this patent.

Claims

1. A passive indoor occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis, characterized in that, By analyzing the amplitude stationarity of CSI data, passive perception and intrusion detection of indoor personnel in various states under normal scenarios can be achieved, including the following steps: 1) Collect Wi-Fi signal channel state information data in indoor environments with and without people to form a sample library; 2) Extract the CSI amplitude matrix from the original sample data and perform denoising and outlier processing using wavelet thresholding and Hample filtering methods; 3) Calculate the global stability GS and local stability LS of the CSI amplitude matrix over a period of time. The global stability GS is the average standard deviation of the amplitudes of N subcarriers within the time period. The local stability LS is the average proportion of the window duration during which K subcarriers are determined to be stable within the time period. The specific process of sliding window steady-state detection is as follows: m Steady-state detection is performed on 30 subcarriers within a duration of seconds using a sliding window method of length W. k The sample data of each subcarrier is divided into a stable duration set Sd k With fluctuation duration set Fd k If window data satisfy < Where r is the threshold parameter for the fluctuation range of the sample standard deviation, this set of data is determined to be steady-state data and incorporated into the steady-state duration set Sd. k In the middle, otherwise it is incorporated into the fluctuation duration set Fd k The window is moved sequentially backward until the steady-state judgment of all time series data is completed; 4) By using a dual-threshold joint judgment algorithm, the global stability and local stability of the current state are compared with the set thresholds to determine whether there is any intrusion or presence of personnel in the current indoor state.

2. The indoor passive occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis as described in claim 1, characterized in that, The Wi-Fi signals are uniformly transmitted by the deployed routers (APs) under the 802.11n protocol, and the Wi-Fi signals collected in the same indoor environment all come from the same router.

3. The indoor passive occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis as described in claim 1, characterized in that, The collected Wi-Fi channel status information is data over a period of time, covering different locations, different times, line-of-sight and non-line-of-sight conditions, and different people's movement states. At the same time, there is no limit to the number of people when data is collected indoors.

4. The indoor passive occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis as described in claim 1, characterized in that, The Channel State Information (CSI) is a three-dimensional complex matrix of M*N*K at each time step, where the first dimension M is the number of transmit antennas, the second dimension N is the number of receive antennas, and the third dimension K is the number of subcarriers.

5. The indoor passive occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis as described in claim 1, characterized in that, The dual-threshold joint determination algorithm combines global stability information GS and local stability information LS to determine the state: the global stability GS of the current state is used as the threshold value. NOW Compared with the first threshold l1, the local stability LS is... NOW Compared with the second threshold l2, if GS NOW < l1 and LS NOW If the value is > l2, it is determined that the room is currently empty; otherwise, it is determined that the room is currently occupied. If the room was previously determined to be empty and is currently determined to be occupied, an intrusion detection alarm is triggered.

6. The indoor passive occupant sensing and intrusion detection method based on CSI amplitude stationarity analysis as described in any one of claims 1 to 5, characterized in that, The method described is applicable to any indoor environment where Wi-Fi signals can be received.

Images