A wireless sensing method based on WiFi multicarrier

Abstract

The application discloses a wireless sensing method based on WiFi multicarrier, and belongs to the wireless sensing field. Three subcarriers with equal frequency difference are selected from collected CSI data, and the first two and the last two subcarriers are divided into two pairs in the order of frequency from small to large. The ratio of the CSI of each pair of subcarriers is calculated. The two ratios are divided again, and the ratio of the two ratios is calculated to obtain a multicarrier secondary ratio. A mapping relationship between the multicarrier secondary ratio and the change in the reflection path length caused by the target motion is constructed to obtain a human motion sensing result. The application is aimed at the problems of amplitude jump and random phase offset in the actual channel state information collected on the single-antenna and multi-antenna wireless sensing system with separate transmitting and receiving. Fine-grained subcarrier information provided by a new generation of WiFi card is used to eliminate the influence of gain control and phase offset through twice division operation, so that ideal CSI is obtained for wireless sensing.

Description

TECHNICAL FIELD

[0001] The application provides a signal preprocessing method for a channel state information (CSI) single-antenna and multi-antenna wireless sensing system based on the separation of transceiver devices, which preprocesses actual CSI and obtains ideal CSI for wireless sensing. Specifically, it relates to a wireless sensing technology based on multi-carrier quadratic ratio, and belongs to the field of ubiquitous computing and wireless sensing. BACKGROUND

[0002] Most current WiFi sensing research mainly uses WiFi cards supporting the 802.11n protocol, such as Intel 5300 cards and Atheros AR9k series cards. With the updating of WiFi technology, new generation WiFi cards supporting 802.11ac / ax have begun to replace 802.11n cards and become more and more common in life. These new generation WiFi cards use wider bandwidth and more advanced technologies, such as multi-user multiple-input multiple-output (MU-MIMO), to provide faster data transmission speeds. However, these new communication characteristics require more fine-grained gain control, which negatively affects WiFi-based wireless sensing. Gain control is a technology that adjusts the energy of WiFi signals, which can adjust the gain of WiFi signals according to the quality of the signal to improve the reliability of communication. However, fine-grained gain control can also cause random signal distortion in the actual CSI signal collected, which is very different from the ideal CSI. Therefore, the performance of some past WiFi sensing methods on these new generation WiFi cards has declined, and sometimes they cannot work. In order to eliminate the influence of gain control, a direct method is to turn off the automatic gain control (AGC), but this will inevitably affect the communication performance. In order to affect the communication performance as little as possible, the best way is to process it through a software method after obtaining the actual CSI. However, different manufacturers have different gain control strategies, and the introduced noise performance will differ. In the absence of prior knowledge (such as noise distribution), it is difficult for traditional filters to completely eliminate these random noises. In particular, in order to support the technical features of the new protocol such as MU-MIMO, each antenna on the new generation card usually uses independent gain control. There are also some transceiver devices with only one antenna. This makes the signal preprocessing technology based on multiple antennas no longer applicable to the new generation of cards. In addition, due to the unsynchronized clocks between transceiver devices, the actual CSI phase also contains a random phase offset, which makes the phase of the actual CSI also cannot be directly used for sensing applications. SUMMARY

[0003] In order to solve the above problems, the present application aims to provide a signal preprocessing method based on WiFi multicarrier, which eliminates the signal distortion caused by gain control and phase offset on actual CSI on the single antenna and multi-antenna wireless sensing system with transceiver separation, so as to obtain ideal CSI for two typical wireless sensing applications.

[0004] The technical scheme provided by the present application is:

[0005] A wireless sensing method based on WiFi multicarrier, the specific steps include:

[0006] S1. Collecting the CSI data of WiFi multicarrier;

[0007] S2. Selecting three subcarriers with equal frequency difference from the collected CSI data, and dividing the first two and the last two subcarriers into two pairs in the order of increasing frequency; calculating the ratio of the CSI of each pair of subcarriers; and then performing a division operation again on the two ratios to calculate the ratio of the two ratios, obtaining the multicarrier secondary ratio;

[0008] S3. Constructing the mapping relationship between the multicarrier secondary ratio and the change of the reflection path length caused by the target motion, that is, when the dynamic path length increases / decreases, the trajectory of the multicarrier secondary ratio on the complex plane rotates clockwise / counterclockwise;

[0009] S4. Obtaining the human motion condition according to the change of the multicarrier secondary ratio.

[0010] Further, one or two pairs of transceiver devices are arranged in step S1 to support different wireless sensing applications, wherein one pair of transceiver devices is required for breath monitoring, and one transmitting device and two receiving devices are required for trajectory tracking, the distance between the two pairs of transceiver devices is equal, and the two pairs of transceiver devices are vertically placed.

[0011] Further, for the application of breath detection, all subcarriers are used to construct a plurality of multicarrier secondary ratio signals, a one-dimensional breath waveform is extracted for each multicarrier secondary ratio signal by using principal component analysis method, and then the breath frequency is estimated by using autocorrelation method on the breath waveform. The accuracy is improved by fusing the breath frequencies estimated from the plurality of multicarrier secondary ratio signals.

[0012] Further, for the application of trajectory tracking, all subcarriers are used to construct a plurality of multicarrier secondary ratio signals, each multicarrier secondary ratio signal is converted to a time-frequency spectrum for analysis, and the frequency component with the maximum energy is selected from the time-frequency spectrum as the Doppler shift, so as to calculate the radial velocity of the target relative to the pair of transceiver devices. The accuracy is improved by fusing the radial velocities estimated from the plurality of multicarrier secondary ratio signals. The real motion speed of the target, including the motion speed and the motion direction, is obtained by combining the radial velocity information of the target relative to the two pairs of transceiver devices.

[0013] Compared with the prior art, the present application has the following advantages:

[0014] 1) The present application uses the fine-grained subcarrier information provided by the new generation of WiFi card, and through two division operations, the influence of gain control and phase offset is eliminated, so that wireless sensing application can be realized by using the CSI of the new generation of WiFi card;

[0015] 2) The technical solution of the present application does not need to close AGC, and does not need to synchronize the clock of the WiFi transceiver device, so it will not affect the communication function of WiFi, and at the same time, it is compatible with wireless sensing and communication functions;

[0016] 3) The technical solution of the present application can simultaneously use the amplitude and phase information of the signal for wireless sensing, improving the accuracy and reliability of wireless sensing;

[0017] 4) The technical solution of the present application can be applied to various wireless sensing applications, such as sensing of small-scale motion and large-scale motion, and has wide application prospects. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 . The flowchart of the WiFi multi-carrier based wireless sensing method of the present application;

[0019] Figure 2 . Amplitude and phase of ideal CSI;

[0020] Figure 3 . Results before and after one division of two carriers, where (a) amplitudes of two subcarriers on one antenna, (b) amplitude ratio of two subcarriers;

[0021] Figure 4 . Results before and after two divisions of multi-carriers, where (a) phase difference between two subcarriers, (b) phase of two divisions of three subcarriers. DETAILED DESCRIPTION

[0022] The present application proposes a WiFi multi-carrier based signal preprocessing method, which will be further described by embodiments in conjunction with the drawings, but does not limit the scope of the present application in any way.

[0023] As shown in Figure 1 , in this embodiment, one pair to two pairs of transceiver devices are arranged to support different wireless sensing applications. Among them, breathing monitoring needs a pair of transceiver devices. And trajectory tracking needs a transmitting device and two receiving devices, the distance between the two pairs of transceiver devices is equal in length, and is placed vertically. A notebook computer is used to receive and process CSI data.

[0024] The actual CSI received can be expressed mathematically as

[0025]

[0026] Where G(t) is the time-varying gain effect, θ o (f, t) is the time-varying phase offset. H(f, t) is the ideal CSI, which includes multiple static paths and dynamic paths caused by human motion. Multiple static paths can be superimposed into a constant H s (f) When the target is far away from the transceiver, the dynamic path caused by the target motion can be approximated as a path in breathing detection and trajectory tracking applications. The ideal CSI can be expressed as

[0027]

[0028] in a(f) is the signal strength of the dynamic path, d0 is the initial length of the dynamic path, and Δd(t) is the change in the dynamic path length over time. When the dynamic path length changes by one wavelength, the ideal CSI will rotate one circle, as shown in Figure 2 As shown, this is a 360-degree rotation in the complex plane. The ideal CSI also exhibits variations in amplitude and phase, exhibiting peaks and valleys. However, the actual received CSI contains time-varying gain effects and phase offsets that mask these phenomena. Fortunately, to achieve high bit rate transmission over frequency-selective fading channels, WiFi employs OFDM signal design. Each WiFi channel is divided into many subcarriers. For example, in 802.11ac, 20 MHz and 40 MHz channels are divided into 64 and 128 subchannels, respectively, referred to as subcarriers. Although the gains vary over time, at any given timestamp, they are identical across different subcarriers. Figure 3 The CSI amplitudes of the two subcarriers are displayed, and it can be seen that they have dramatic signal jumps over time.

[0029] The present invention divides the actual CSI of the two subcarriers to eliminate the signal distortion introduced by the gain control. The ratio of the two subcarriers can be expressed as

[0030]

[0031] Although a single division operation can eliminate the signal distortion introduced by gain control, the phase of the CSI ratio of the two subcarriers still contains non-negligible deviation. This is because although the two subcarrier CSIs share the same antenna gain, the random phase offsets of different subcarriers are not exactly the same. Therefore, the phase offset cannot be completely eliminated by dividing the signal between the two subcarriers. Figure 4 (a) shows Figure 3The residual phase offset is contained in the phase of the ratio of two subcarrier signals. Considering that the CSI phase offset is linearly related to the subcarrier frequency, the present invention introduces an additional subcarrier to handle the residual phase offset. Specifically, the present invention selects two pairs of subcarriers with equal frequency intervals, and the residual phase offset in the ratio of the two pairs of subcarriers is the same. Therefore, the present invention eliminates the residual phase offset by performing a division operation again on the two ratios. Compared with a single division, the present invention introduces a third subcarrier with a frequency of f+2Δf, which forms two pairs of subcarriers with the same frequency difference (the first pair: f and f+Δf, and the second pair: f+Δf and f+2Δf). Then the multicarrier second ratio can be defined as

[0032]

[0033] In equation (4), θ o (f, t) - θ o (f+Δf, t) = θ o (f+Δf, t) - θ o (f+2Δf, t). From Figure 4 (b) it can be seen that the residual phase offset can be eliminated by two division operations. In order to further derive the relationship between the multicarrier second ratio and the change in the reflection path length caused by human motion, the present invention now applies the following steps:

[0034] · This is because the frequency difference between two subcarriers (i.e. a few MHz) is very small compared with the subcarrier frequency (e.g. 5 GHz) which is at the level of gigahertz. If the path length change Δd(t) is λ, then the dynamic path phase change of all subcarriers within the narrow bandwidth is almost 2π. Therefore, the present invention can use a uniform symbol to represent and When Δd(t) increases by λ, a unit circle that rotates clockwise in the complex plane.

[0035] · The present invention lets and represent H s (f), H s (f+Δf) and H s (f+2Δf), respectively. The present invention further lets and represent A(f), A(f+Δf) and A(f+2Δf), respectively.

[0036] In this way, the multicarrier second ratio can be rewritten as

[0037]

[0038] In wireless sensing, dynamic component refers to the signal reflected by moving targets. Compared with LoS (Line-of-Sight) path signal, dynamic component is weaker because dynamic path propagates longer distance and the received dynamic signal energy at the receiving device is lower. Therefore, the product of two dynamic components (i.e. ③ and ⑥) is much smaller than the product of two static components or the product of a dynamic component and a static component, and can be ignored. The product of static components of two subcarriers (i.e. ① and ④) can be regarded as a constant. Through the above analysis, the calculated multi-carrier quadratic ratio of the present application is in the form of a Mobius transform. Since the Mobius transform still maintains the property, the present application constructs the mapping relationship between the multi-carrier quadratic ratio and the change of the reflection path length caused by target movement, i.e. when the dynamic path length changes by one wavelength, the trajectory of the change of the multi-carrier quadratic ratio on the complex plane is still a circle. When the path length change is less than one wavelength, the trajectory is an arc. In actual sensing applications, the rotation direction of the multi-carrier quadratic ratio is the same as that of the dynamic path length, i.e. when the dynamic path length increases / decreases, the trajectory of the change of the multi-carrier quadratic ratio on the complex plane presents clockwise / anticlockwise rotation.

[0039] Therefore, the present application provides a wireless sensing method based on WiFi multi-carrier, and the specific steps include:

[0040] S1. According to the wireless sensing application of different movements of human body, collect the CSI data of WiFi multi-carrier;

[0041] S2. Select three subcarriers with equal frequency difference from the collected CSI data, and divide the first two and the last two subcarriers into two pairs in the order of increasing frequency. Calculate the ratio of each pair of subcarrier CSI. Then, perform a division operation again on the two ratios to calculate the ratio of the two ratios, and obtain the multi-carrier quadratic ratio;

[0042] S3. Construct the mapping relationship between the multi-carrier quadratic ratio and the change of the reflection path length caused by target movement, i.e. when the dynamic path length increases / decreases, the trajectory of the change of the multi-carrier quadratic ratio on the complex plane presents clockwise / anticlockwise rotation;

[0043] S4. Obtain the sensing result of human body movement according to the change of the multi-carrier quadratic ratio.

[0044] In order to make full use of the diversity of subcarriers, the present application processes all subcarriers to improve the robustness of the system. Specifically, the present application uses all subcarriers to construct multiple groups of multicarrier quadratic ratios. For example, a WiFi card can provide CSI information of 52 subcarriers, then the 1st, 18th and 35th subcarriers can be selected as a group, the 2nd, 19th and 36th subcarriers can be selected as another group. In this way, the present application can obtain a total of 17 groups of subcarriers for twice division operation.

[0045] In the application of the present application in breath detection, based on the property derived from equation (5), the trajectory of the change of the multicarrier quadratic ratio signal on the complex plane is an arc of a circle because the change of the reflection path length caused by the chest fluctuation during each inhalation or exhalation is less than a wavelength. And the rotation directions of the trajectories of the corresponding multicarrier quadratic ratio signals on the complex plane during exhalation and inhalation are opposite. Based on this phenomenon, the present application first applies principal component analysis method to the multicarrier quadratic ratio signal to convert the two-dimensional multicarrier quadratic ratio signal on the complex plane into a one-dimensional respiratory waveform, which reflects the pattern of the change of the reflection path length caused by human respiration. Then, the present application applies Savitzky-Golay filter to smooth the extracted respiratory waveform. After that, the present application estimates the respiratory frequency by using autocorrelation method for the smoothed respiratory waveform data, in which the window size is set to 30 seconds. In order to improve the accuracy of the results, the present application estimates the respiratory frequency by using the multicarrier quadratic ratio signals of different carrier combinations, and obtains the final respiratory frequency by averaging all respiratory frequency estimation results.

[0046] In the application of trajectory tracking, three devices are needed, including a transmitting device and two receiving devices, which are denoted as rx1 and rx2 respectively, and the line connecting the transmitting device and rx1 is perpendicular to the line connecting the transmitting device and rx2. The key information in the application of human body trajectory tracking is the Doppler shift, from which the movement speed of the target can be solved. In combination with the geometric position relationship of the target and the transmitting and receiving devices, the change speed of the dynamic path length can be converted into the radial speed of the target relative to the transmitting and receiving devices. With another pair of transmitting and receiving devices, the radial speed in another direction can be obtained. The two radial speeds are integrated to obtain the actual speed of the target. Through the mapping relationship between the constructed multi-carrier second ratio and the change value of the reflection path length caused by the human body movement, the Doppler shift of the target relative to the transmitting and receiving devices can be estimated. Specifically, the multi-carrier second ratio signal is converted to the time-frequency spectrum for analysis by using continuous wavelet transform (CWT). The frequency component with the maximum energy is selected from the time-frequency spectrum as the Doppler shift, so as to calculate the radial speed of the target relative to a pair of transmitting and receiving devices. In combination with the radial speed information of the target relative to two pairs of transmitting and receiving devices, the real movement speed of the target, including the movement speed and the movement direction, can be obtained. On the premise of given initial position, the trajectory of the target can be constructed by continuously estimating the speed of the target. Meanwhile, Kalman filtering is used to further smooth the trajectory.

[0047] The ideal CSI can be obtained by preprocessing the actual CSI by using the application, so that the application can be applied to various wireless sensing applications, such as gesture recognition, fall detection, etc., and has wide application prospects.

[0048] It should be noted that the purpose of publishing the embodiments is to help further understand the application, but those skilled in the art can understand that various replacements and modifications are possible without departing from the spirit and scope of the application and the appended claims. Therefore, the application should not be limited to the disclosed content of the embodiments, and the scope of protection claimed by the application is subject to the scope defined by the claims.

Claims

1. A method of wireless sensing based on WiFi multicarrier, characterized in that, The specific steps include: S1. Collecting CSI data of WiFi multicarrier; S2. Selecting three subcarriers with equal frequency difference from the collected CSI data, and dividing the first two and the last two subcarriers into two pairs in the order of increasing frequency; calculating the ratio of the CSI of each pair of subcarriers; and performing a division operation again on the two ratios to calculate the ratio of the two ratios, thereby obtaining a multicarrier secondary ratio; S3. Constructing a mapping relationship between the multicarrier secondary ratio and the change in the length of the reflection path caused by the target motion, that is, when the dynamic path length increases / decreases, the trajectory of the multicarrier secondary ratio on the complex plane rotates clockwise / counterclockwise; S4. Obtaining the human motion condition according to the change in the multicarrier secondary ratio, which includes a breathing detection application and a trajectory tracking application; for the breathing detection application, all subcarriers are used to construct multiple multicarrier secondary ratio signals, and a one-dimensional breathing waveform is extracted from each multicarrier secondary ratio signal by using a principal component analysis method; then, a breathing frequency is estimated by using an autocorrelation method on the breathing waveform, and the breathing frequencies estimated from multiple multicarrier secondary ratio signals are fused; for the trajectory tracking application, all subcarriers are used to construct multiple multicarrier secondary ratio signals, each multicarrier secondary ratio signal is converted to a time-frequency spectrum for analysis, the frequency component with the maximum energy in the time-frequency spectrum is selected as the Doppler shift, thereby calculating the radial velocity of the target relative to a pair of transceiver devices, the radial velocities estimated from multiple multicarrier secondary ratio signals are fused, and the real motion speed of the target, including the motion rate and the motion direction, is obtained by combining the radial velocity information of the target relative to two pairs of transceiver devices.

2. The WiFi multicarrier based wireless aware method of claim 1, wherein, In step S1, one pair to two pairs of transceiver devices are arranged to support different wireless sensing applications, wherein breathing monitoring requires one pair of transceiver devices, and trajectory tracking requires one transmitting device and two receiving devices; the distance between the two pairs of transceiver devices is equal, and they are placed vertically.

3. The WiFi multicarrier based wireless aware method of claim 1, wherein, The Savitzky-Golay filter is used to smooth the extracted breathing waveform, and then the autocorrelation method is used on the smoothed breathing waveform data.

Images