Long-distance human respiration frequency evaluation method based on LoRa signal

Abstract

The invention belongs to the field of wireless sensing, and relates to a long-distance human respiration frequency evaluation method based on a LoRa radio frequency signal, and the method comprises the following three steps: (1) data collection: covering the position of a human body through a radio frequency signal, and collecting reflected signals of different individuals at different positions and orientations; and (2) data processing and enhancement: eliminating random phase noise in the signals by carrying out ratio operation on the signals, carrying out denoising and smoothing processing by using a filter, realizing data enhancement through signal rotation, and finally generating a time-frequency diagram by adopting short-time Fourier transform. And (3) constructing the relationship between the signal change and the breathing frequency: based on the generated time-frequency graph, establishing a relationship model between the signal change and the breathing frequency during detection by utilizing a visual model. According to the invention, remote detection of the breathing frequency of the user can be realized under the condition that the user is not interfered, so that an effective basis is provided for scenes such as intrusion detection, disaster rescue and the like.

Description

Technical Field

[0001] The present invention belongs to the field of wireless sensing, and specifically relates to a long-distance human breathing rate assessment method based on LoRa radio frequency signals, which is mainly used to solve the problem that the existing wireless sensing technology has a small sensing range and is difficult to accurately measure the target user's breathing rate in long-distance and wall-penetrating scenarios. Background Art

[0002] Human vital signs detection has been widely used in a variety of scenarios such as energy management systems in buildings, intruder detection, and disaster relief. Accurate human vital signs detection provides a reliable basis for scenarios such as security and emergency rescue. There are currently a variety of methods to achieve non-contact human vital signs detection, including vision, sensors, and radio frequency signals. Vision-based systems use a large number of cameras to identify human targets, but face problems such as light, line of sight, and privacy protection. Sensor-based methods require expensive equipment and regular maintenance, which increases the cost of use. In contrast, radio frequency signals can achieve non-line-of-sight human target detection, which can achieve high coverage in practice and achieve a good balance between accuracy and privacy protection. However, existing radio frequency signal-based methods, such as WIFI, RFID, Bluetooth, and acoustic technologies, have a key limitation: the sensing distance is limited, and wireless sensing can usually only be achieved within a few meters. This makes them only able to be used for vital signs detection in a small range, which is difficult to meet the needs of some practical application scenarios.

[0003] LoRa is a low-power wide area network communication technology that uses linear frequency modulation spread spectrum (CSS) technology to significantly increase the communication distance while maintaining low power consumption. In suburban environments, its transmission distance can reach several thousand meters. In addition, LoRa devices are inexpensive and have assumed the function of data transmission in fields such as environmental monitoring, factory management, and smart agriculture and animal husbandry. Compared with signals such as WIFI and RFID, LoRa has less loss during transmission and is more suitable for remote perception of human behavior. Therefore, using LoRa to achieve long-distance human vital signs detection has strong feasibility and high application value. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a long-distance human breathing rate assessment method based on LoRa signals, which can realize the detection of the breathing rate of human targets in long-distance and through-wall scenarios.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A remote human respiratory rate assessment method based on LoRa signals, characterized in that it comprises the following steps:

[0007] 1) The data collection process is as follows:

[0008] The LoRa node cooperates with the directional antenna to transmit the LoRa radio frequency signal S(t) to the area to be detected;

[0009] The target is asked to sit or stand in a sector area based on the perpendicular bisector of the two receiving antennas. The area ranges from a horizontal angle of -35° to 35° and is 5 to 15 meters in front of the antennas.

[0010] Start the LoRa gateway, use two directional antennas to collect the target reflection signals R1(t) and R2(t), and downsample the signals to reduce the time cost of subsequent data processing and enhancement;

[0011] 2) The data processing and enhancement process is as follows:

[0012] The signals collected by the two directional antennas are ratioed to obtain the signal ratio S(t) = R1(t) / R2(t) to eliminate the random phase offset in the received signal and ensure the availability of signal phase information;

[0013] The obtained signal ratio is applied to a low-pass filter to remove high-frequency noise caused by the environment and multipath;

[0014] The signal ratio was smoothed using Savitzky Golay filter;

[0015] The signal ratio is rotated at a fixed step angle on the complex plane to obtain a candidate set of signal ratios to achieve data enhancement;

[0016] The signal ratio candidate sets are sequentially subjected to short-time Fourier transform to obtain the time-frequency graph corresponding to each signal ratio candidate, and the time-frequency graphs are combined into a set of time-frequency graphs.

[0017] 3) Establish the relationship between signal change and breathing rate:

[0018] The time-frequency graph set obtained after data processing and enhancement is used as the input layer variable of the visual model, the number of breaths is used as the output layer variable of the model, and the target's true value of the number of breaths is used as the label;

[0019] Set the model input layer, output layer, adapter training parameters and network structure, and use the input layer variable data and the target's true breathing number to train the input layer, output layer and adapter parameters to fine-tune the visual model;

[0020] The fine-tuned network model is used as the relationship model between signal changes and breathing rate.

[0021] To optimize the above technical solutions, the specific measures taken also include:

[0022] Furthermore, the specific steps of data collection are as follows:

[0023] S1: Pair the LoRa node with a directional antenna and configure it to continuously transmit the signal S(t)=exp{j2πf c t+jπk 2}, where f c is the center frequency, k=B / T is the scanning rate of the linear frequency modulation pulse, B is the frequency bandwidth, and T is the duration of the linear frequency modulation pulse;

[0024] S2: Let the target sit or stand in the range of -35° to 35° in front of the LoRa transmitting antenna, and the distance from the transmitting antenna is 5-15m;

[0025] S3: Building a LoRa gateway based on USRP B210 and GNU Radio, using two LoRa receiving antennas to continuously receive the reflected signal of the target during detection and Among them, θ c is the carrier frequency deviation caused by the inconsistency between the transmitting node and the receiver clock, θ s is the phase error caused by sampling frequency deviation, τ n (t) is the propagation delay of the nth path;

[0026] S4: downsample the signal and save it.

[0027] Furthermore, the specific steps of data processing and enhancement are as follows:

[0028] S1: From the vector point of view, the received signal consists of static vectors and dynamic vectors. The static vector can be approximated as a single static vector in a finite time, and the dynamic vector can be regarded as a function of the target moving distance. Therefore, the received signal can also be expressed as and Where d(t) is the length of the dynamic path caused by the target movement;

[0029] S2: Since the two antennas share the same clock, they have the same carrier frequency offset and sampling frequency offset. The frequency offset can be eliminated by performing a ratio operation on the signals received by the two antennas to obtain the signal ratio:

[0030] Where Δs is the difference in dynamic path length caused by the different positions of the two receiving antennas. Compared with d(t), it can be regarded as a constant in a short time. The signal ratio obtained can be regarded as the Möbius transformation of the signal received by one antenna. This transformation does not change the shape and direction of the signal trajectory in the complex plane, but only changes the position and size of the trajectory. Therefore, the signal ratio still contains the target motion information;

[0031] S3: Determine the frequency range according to the human breathing characteristics and apply a low-pass filter to the signal ratio to remove high-frequency noise caused by the environment and multipath effects;

[0032] S4: Apply Savitzky Golay filter to further smooth the signal ratio;

[0033] S4: The signal is stored in complex form, and a single sampling point can be expressed as: z = a + bi or Ae iφ , where a and b are the real and imaginary parts respectively, and A and φ are the amplitude and phase. When the signal ratio is expressed in terms of real and imaginary parts, the real and imaginary parts are orthogonal to each other, and the two show a complementary trend, that is, when the real component is not conducive to wireless perception at a certain position, the imaginary component is conducive to perception, and vice versa. Therefore, the real and imaginary components can be combined for complementary perception. In order not to be limited to the two dimensions of the real and imaginary parts of the signal, the signal ratio is data enhanced, and the signal ratio is rotated in the complex plane to generate multiple groups of signal ratio candidates. The original signal ratio is stored as Z = (z1, z2, ..., z n ), and rotate it as a whole by θ to obtain a signal ratio candidate: Z′=(z′1,z′2,...,z′ n )=(z1·e iθ , z2·e iθ , ..., z n ·e iθ )

[0034] Vary θ from 0 to 2π with a fixed step size to generate different signal ratio candidates;

[0035] S5: In order to retain the time and frequency information, short-time Fourier transform is used to extract the time-frequency features of the real and imaginary parts of the signal ratio candidates one by one, and the above features are presented in the form of a set of time-frequency diagrams through visualization methods.

[0036] Furthermore, the relationship between the signal change and the number of breaths is constructed as follows:

[0037] S1: The set of time-frequency graphs obtained by data processing and enhancement is used as the input layer variable of the fine-tuning visual model, the number of breaths is used as the output layer variable, and the true value of the number of breaths is used as the label required for model fine-tuning;

[0038] S2: Add low-parameter adapters to the visual model backbone network according to the usage scenario, and freeze other parameters of the backbone network except the adapters;

[0039] S3: The measured target counts the number of breaths by itself as the true value, and the set visual model is fine-tuned using the set of time-frequency graphs corresponding to the signal ratio and the true value of the number of breaths;

[0040] S4: The fine-tuned visual model is used as a relationship model between signal changes and breathing rate. The number of breathing rates of the target during the detection period can be obtained by inputting a set of corresponding time-frequency graphs of the signal ratio.

[0041] Furthermore, in step S3, the number of model training cycles is set to 200, the learning rate is set to 0.001, and the iteration is stopped if the root mean square error of the validation set does not decrease for 10 consecutive times during the model fine-tuning iteration process.

[0042] The beneficial effects of the present invention are as follows: a method for evaluating the number of human breaths at a distance based on a LoRa signal is designed, wherein a LoRa signal is transmitted to the area to be detected, and two receiving antennas are used to continuously receive the reflected signal, and the carrier frequency deviation and sampling frequency deviation of the received signal are eliminated by performing ratio operations on the signals received by the two receiving antennas, and a filter is used to denoise and smooth the signal ratio, and the signal ratio is rotated at the same time to obtain a signal ratio candidate with better perception ability, and then a time-frequency graph set corresponding to the signal ratio candidate is generated by using a short-time Fourier transform and a visualization method, and finally a visual model is fine-tuned with the time-frequency graph set as input and the true value of the number of breaths as a label, thereby establishing a relationship model between signal changes and the number of breaths, and realizing an accurate evaluation of the target number of breaths. The present invention can continuously detect the number of user breaths at a distance without interfering with the user, and provide an effective basis for scenes such as intrusion detection and disaster relief. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Framework diagram of the long-distance human breathing rate assessment method based on LoRa signal

[0044] Figure 2 Diagram for fine-tuning adapter-based vision model parameters Specific implementation methods

[0045] The present invention will now be described in further detail with reference to the accompanying drawings.

[0046] like Figure 1The main idea of ​​the long-distance human breathing rate assessment method based on LoRa signals shown in the figure is to divide the method into three steps: data collection, data processing and enhancement, and the relationship between signal changes and breathing rate. The first two steps collect the reflected signals of the targets in the detection area through LoRa nodes and LoRa gateways, and pre-process the signal data through signal division, low-pass filter, and Savitzky Golay filter. At the same time, the signal is rotated to generate signal ratio candidates, and finally the visualization of time-frequency features is realized through visualization methods. The third step is to establish a relationship model between signal changes and breathing rate during detection by fine-tuning the visual model.

[0047] The specific steps of data collection are as follows:

[0048] 1) Place two receiving antennas and one transmitting antenna at the same vertical height (1.2m) and point them toward the area to be detected. Connect the LoRa node to the transmitting antenna and use a 10000mAh mobile power supply to power the LoRa node.

[0049] 2) The target user to be detected sits or stands in a sector area based on the perpendicular bisector of the two receiving antennas. The area ranges from a horizontal angle of -35° to 35° and is 5 to 15 m in front of the antenna.

[0050] 3) Connect a PC equipped with an Inter Core i7-4790 CPU and 12GB memory to USRPB210 via a USB interface, and connect the two RX interfaces of USRP B210 to two receiving antennas respectively. At the same time, draw a data flow graph based on GNU Radio, and collect the reflected signal data of the area to be detected based on the flow graph, downsample the signal, and save the signal data in the form of real and imaginary parts;

[0051] The specific steps of data processing and enhancement are as follows:

[0052] 1) Since the transmitting node and the receiver clock are not synchronized, the received signal has a phase offset caused by the carrier frequency offset and the sampling frequency offset. The received signals of the two antennas are expressed as and From a vector perspective, the two receiving antennas can be represented as and Where d(t) is the length of the dynamic path caused by the target movement. By dividing the signals received by the two antennas, the phase offset in the signal can be eliminated, and the resulting signal ratio is:

[0053] Compared with d(t), Δs caused by the different positions of the two receiving antennas can be regarded as a constant. This signal ratio can be regarded as the Mobius transformation of the signal received by a single receiving antenna under ideal conditions. This transformation does not change the shape and direction of the original signal trajectory on the complex plane, but only changes the position and size of the signal trajectory. The target motion information is still included in it;

[0054] 2) Based on the upper limit of normal human breathing rate, the theoretical maximum frequency of the received signal ratio is calculated, and a low-pass filter is applied to the signal ratio to remove the high-frequency noise generated by the environment and multipath effects;

[0055] 3) Based on the low-pass filtering, the signal ratio is further smoothed using the Savitzky Golay filter;

[0056] 4) The signal ratio is composed of multiple sampling points, each of which is stored in a complex value form. Each sampling point in the signal ratio is rotated by an angle θ in turn, thereby rotating the signal ratio by θ. The result of the rotation is a signal ratio candidate. Multiple signal ratio candidates are generated by changing θ with a fixed step size from 0 to 2π;

[0057] 5) Use short-time Fourier transform to extract time-frequency features from the real and imaginary parts of the signal ratio candidates one by one to present time and frequency information simultaneously. At the same time, the time-frequency features are saved as a set of time-frequency graphs through visualization technology.

[0058] The specific steps to construct the relationship between signal changes and breathing rate are:

[0059] 1) Build a deep learning Linux development environment, based on the deep learning framework Pytorch, adjust the input layer and output layer of the visual model to adapt to the expected input and output. Figure 2 The structure shown is to add a low-parameter adapter inside the visual model;

[0060] 2) During the test, the target records the number of breaths by itself as the true value of the number of breaths. The true value and the set of time-frequency graphs obtained by data processing and enhancement together constitute a data sample, where the former is the label and the latter is the input.

[0061] 3) Divide the dataset into training dataset, test dataset, and validation dataset in the ratio of 7:2:1:

[0062] 4) Fine-tune the set visual model using the set of time-frequency graphs corresponding to the signal ratio candidates and the true value of the number of breaths. At the same time, update the parameters of the input layer, adapter, and output layer according to the Adam optimization algorithm of Pytorch. The number of cycles is 200, the learning rate is set to 0.001, and the root mean square error (RMSE) of the validation set does not decrease for 10 consecutive times during the model fine-tuning iteration process, then stop the iteration;

[0063] 5) The fine-tuned visual model is used as a relationship model between signal changes and breathing rate, and a set of time-frequency graphs corresponding to the signal ratio obtained from the target reflection signal is input to obtain the target breathing rate during the detection period.

[0064] It should be noted that the terms such as "upper", "lower", "left", "right", "front", "back", etc. cited in the invention are only for the convenience of description and are not used to limit the scope of implementation of the present invention. Changes or adjustments in their relative relationships should be regarded as the scope of implementation of the present invention without substantially changing the technical content.

[0065] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A long-distance human respiratory rate assessment method based on LoRa signals, characterized in that: The steps include: 1) Data collection: Use LoRa nodes with directional antennas to transmit LoRa radio frequency signals to the area to be detected; Use USRP B210 to connect two directional antennas to continuously receive the reflected signals from the target area, which are represented as R1(t) and R2(t) respectively. The received signals are downsampled and saved. 2) Data processing and enhancement: The signal ratio S(t) = R1(t) / R2(t) is obtained by performing ratio calculation on the signals collected by the two receiving antennas, eliminating the random phase offset of the signal caused by the asynchrony between the transmitting node and the receiver clock, and restoring the availability of the signal phase information; The upper limit of the signal ratio frequency is calculated based on the theoretical breathing rate of the human body, and a low-pass filter is used to remove the high-frequency noise caused by the environment and multipath effects; The filtered signal is further smoothed using a Savitzky Golay filter; The processed signal ratio is rotated step by step from 0 to 2π with a fixed step size to obtain multiple signal ratio candidates; Applying short-time Fourier transform to the signal ratio candidates one by one to obtain the corresponding time-frequency graphs, and further combining them into a set of time-frequency graphs; 3) Establish the relationship between signal change and breathing rate: The time-frequency graph set obtained by short-time Fourier transform is used as the input layer variable of the visual model, the breathing number is used as the output layer variable, and the true value of the breathing number is used as the label; Adjust the model input layer and output layer to adapt to the input and output variables, add a low-parameter adapter to the visual model backbone network and freeze the remaining parameters, use the input layer variable data and the corresponding true value of the number of breaths to train the input layer, output layer and low-parameter adapter parameters to achieve fine-tuning of the visual model; The fine-tuned visual model is used as the relationship model between signal changes and breathing rate.

2. The method according to claim 1, characterized in that: The specific steps of data processing and enhancement are as follows: S1: The received signal can be regarded as composed of static vectors and dynamic vectors. The static vector can be regarded as a single static vector in a short time. Therefore, the signals received by the two receiving antennas can be expressed as and where d t (t) is the length of the dynamic path caused by the target movement; S2: Two antennas are connected to the same USRP and share the same clock. Therefore, the two antennas have the same carrier frequency offset and sampling frequency offset. The signal ratio obtained by performing ratio calculation on the signals received by the two antennas can eliminate the frequency offset in the original signal and ensure the availability of phase information: Where Δs is the difference in dynamic path length caused by the different positions of the two receiving antennas. Compared with d(t), it can be regarded as a constant in a short time. Therefore, the signal ratio can be regarded as the Mobius transformation of the signal received by one antenna. This transformation will not change the shape and direction of the signal trajectory on the complex plane, but only the position and size of the trajectory. The target motion information is still included in it. S3: Determine the theoretical frequency range of the signal based on the human breathing characteristics, apply a low-pass filter to the signal ratio to remove the high-frequency noise caused by the environment and multipath effects in the signal ratio, and then use a Savitzky Golay filter to further smooth the signal ratio; S4: The signal ratio is composed of multiple sampling points, each sampling point is represented by a complex value, and for each sampling point and e iθ By performing multiplication operation, the sample points can be rotated by θ, and the sample points in the signal ratio can be rotated one by one to achieve the overall signal ratio rotation by θ, thus generating a signal ratio candidate; S5: changing θ with a fixed step length from 0 to 2π to generate multiple signal ratio candidates; S6: Use short-time Fourier transform to extract time-frequency domain features from the real part and imaginary part of the signal ratio candidate in turn, present the time-frequency features in the form of time-frequency graphs through visualization methods, and further combine the obtained time-frequency graphs into a set of time-frequency graphs.

3. The implementation method according to claim 1, characterized in that: The specific steps for constructing the relationship between the signal change and the number of breaths are as follows: S1: The time-frequency graph set obtained after data processing and enhancement is used as the input layer variable of the visual model, the number of breaths is used as the output layer variable, and the true value of the number of breaths is used as the label required for model fine-tuning; S2: Adjust the input and output layers of the visual model according to the data format, add low-parameter adapters to the backbone network of the visual model, and then freeze the remaining parameters of the backbone network except the adapters; S3: The measured target counts the number of breaths by itself as the true value, and the set visual model is fine-tuned using the set of time-frequency graphs corresponding to the signal ratio and the true value of the number of breaths; S4: The fine-tuned visual model is used as a relationship model between signal changes and breathing rate, and the target breathing rate during the detection period can be obtained by inputting a set of corresponding time-frequency graphs of the signal ratio.

4. The method for evaluating the number of respirations according to claim 3, wherein: In step S3, the number of fine-tuning cycles of the visual model is set to 200, the learning rate is set to 0.001, and the iteration is stopped if the root mean square error of the validation set does not decrease for 10 consecutive times during the model fine-tuning iteration.

Images