A passive wireless sensor signal anti-interference method based on RFID

3.1 The anti-interference method for sensor signals based on RFID technology

RFID is a wireless communication technology that identifies and tracks tags attached to items through radio waves. The RFID can quickly read data without the need for connecting antennas between tags and readers. This makes RFID a highly effective automatic recognition and data capture technology [13]. The basic framework of RFID is shown in Figure 2.

Figure 2:

Basic framework of RFID. Note: Figure 2 shows the core components of an RFID system, including signal data management, signal data reading, and electronic labels. This diagram can analyze signal data through the network and transmit the data to the electronic label through antennas, which is the foundation for understanding how RFID technology operates.

In Figure 2, RFID technology consists of three main parts: signal data management, signal data reading, and electronic labels. The signal data are analyzed by the reader through the network, and then transmitted to the electronic label through the reader’s antenna [14], 15]. The general RFID system is divided into active and passive types. Passive generally refers to the system being able to independently read and write energy without providing signal transmission process. Passive is analyzed. When analyzing infinite passive sensor signals, passive systems are selected. Passive systems are also divided into inductive coupling and electromagnetic coupling [16], 17]. The RFID system with inductive coupling is shown in Figure 3.

Figure 3:

The RFID system with inductive coupling. Note: Figure 3 illustrates the workflow of the RFID system with inductive coupling, which involves generating a magnetic field through a power supply, converting the magnetic signal into an electrical signal through magnetic induction, and finally transmitting it to an electronic label. This helps to understand the energy transfer and signal conversion in RFID systems.

In the system shown in Figure 3, the RFID reader can first generate a magnetic field through the power supply, and then increase this induction method through magnetic induction. Afterwards, the magnetic signal is converted into an electrical signal through magnetic induction and transmitted to the electronic label. During this process, modulation is carried out according to relatively ordered signal control and phase frequency, which transforms the frequency of the entire system into a load that can be added. Usually, the system working in this way can achieve more efficient processing results. The reflection modulation system is shown in Figure 4.

Figure 4:

Reflection modulation system. Note: Figure 4 illustrates the modulated RFID system and explains the reflection, scattering, and diffraction phenomena encountered by wireless passive sensors when propagating in air, as well as how these phenomena affect signal propagation and reception.

The modulated RFID system shown in Figure 4 consists of components such as a reader, directional coupler, signal transmission, electronic label, and antenna. The signal transmission process of the passive RFID system is achieved through signal transmission of different components. When wireless passive sensors propagate in the air, they usually encounter three situations: reflection, scattering, and diffraction, which not only include large-scale signal propagation, but also small-scale signal fading and transmission. In signal propagation, interval code interference may occur due to the frequency and scattering of the signal, as shown in equation (1) [18].

In equation (1), h t represents the impulse response. n R t represents the output noise generated by filtering additive noise. a _k represents the transmission sequence signal. By sampling the signal sequence, the interval interference shown in equation (2) is obtained [19].

In equation (2), a k h j − k T + γ 0 + n g j T + γ 0 represents the sum of sampling moments for all code regions. jT + γ ₀ represents the moment when the sample is sampled. Other parameter expressions are the same as above. Due to the limited efficiency of RFID systems in signal reception and transmission, interval code interference usually only focuses on signal fading. To reduce the current signal degradation, a new improved blind equalization algorithm is added. By calculating the minimum extremum point of the current system, the ideal system construction process for signal attenuation in the current interval interference model is achieved, reducing the impact of interval interference. Equation (3) shows the training sequence process of blind equalizer.

In equation (3), s ̂ ( n ) represents the estimated signal. g ⋅ represents a non-linear function without memory. x ⌣ ( n ) represents the transmitted signal. The error size of the current estimated value is shown in equation (4).

In equation (4), e(n) represents the size of the error signal. The other parameters are the same as the above expression. The algorithm obtained from the error data is shown in equation (5).

In equation (5), W(n) represents the transverse filter signal. μ represents the step size factor. Y(n) represents the input signal. The other parameters are the same as above. By calculating the convergence, the anti-interference ability of the current algorithm is analyzed, as shown in equation (6).

In equation (6), E e n y n − i represents the convergence situation and convergence conditions of the algorithm. i represents the variable size. The remaining parameters are the same as above. The two sides of equation (6) are simultaneously multiplied by w _i−k (n − k). The variables are summed to obtain a convergence equation, as shown in equation (7).

In equation (7), k represents the size of the new variable. L represents the size of the interference signal. When the L value is large enough, the entire equation can be approximately transformed into equation (8).

The parameter expression in equation (8) is the same as the above equation. Finally, equation (3) is combined with equation (8) to obtain the final expression, as shown in equation (9).

The parameter expression in equation (9) is the same as the above equation. When the sequence output by the algorithm comes from the same nonlinear function, the entire process can be satisfied by the blind equalization algorithm. The anti-interference ability of the sensor signal can be enhanced by calculating the minimum extremum. Meanwhile, sensor identity signal authentication is also a way to improve the anti-interference ability of sensor signals.

In RFID systems, signals may be affected by multipath effects and EMI during transmission, resulting in signal distortion. The blind equalization algorithm achieves signal anti-interference through signal reception, signal pre-processing, signal equalization, and signal identification to extract useful information. The mathematical model and processing flow of the blind equalization algorithm are described in Equations (1)– (9), respectively. This includes parameters such as the impulse response of the signal, output noise, transmission sequence signal, and how the equalization process is optimized through sampling, error calculation, and algorithm convergence. The blind equalization algorithm is able to adaptively adjust its parameters to changes in channel conditions. Meanwhile, based on the blind equalization algorithm, the new method significantly improves the anti-interference capability of the RFID system and achieves higher anti-interference performance, which is especially important for the application of RFID systems in IIoT.

3.2 RFID-based sensor signal frequency authentication and recognition

By incorporating frequency authentication and data recognition, the anti-interference effect of current sensors and the security of signal data can be improved. In wireless signal transmission and reception of sensors, device reception authentication is a means to ensure the security of sensor wireless signals [20], 21]. Nowadays, wireless signals are frequent. Excessive and redundant signal information can easily cause sensor failure or even reception errors. Therefore, incorporating RFID identity authentication and recognition technology into the signal transmission process of sensors can effectively improve the usage time and signal reception effect of passive wireless sensors. The digital signal transmission framework of wireless passive sensors is shown in Figure 5.

Figure 5:

Wireless passive sensor digital signal transmission framework. Note: Figure 5 illustrates the process of transmitting wireless passive sensor signals from the signal processing interface at the sensor front-end to the RF front-end, which is crucial for understanding the digital signal processing of sensor signals.

In Figure 5, after the wireless passive sensor signal is transmitted, it first needs to be transmitted to the radio frequency front-end through the signal processing interface of the sensor front-end, including some signal transmission structures such as data model converter, IF filter converter, frequency conversion structure device, and RF filter converter. After amplifying and processing the signal, the digital signal is converted into a high-frequency signal. Finally, the signal is transmitted out through the antenna device. When authenticating sensor frequency signals, there are four steps involved: signal acquisition, signal preprocessing, signal feature extraction, and signal authentication. The signals emitted by the machine are collected, followed by normalization and noise reduction for preprocessing analysis. Machine learning methods are used to extract and validate features of signals through email and unsupervised methods. Finally, the signal reception analysis is completed. The signal extraction process framework is shown in Figure 6.

Figure 6:

Framework of signal extraction process. Note: Figure 6 illustrates the application of machine learning in receiving machine signals. This framework improves the security performance and anti-interference ability of current signals through supervised learning and frequency authentication of RF signals.

In Figure 6, the supervised learning radio frequency authentication block diagram in receiving machine signals has an additional sensor identity authentication and recognition function compared with the unsupervised block diagram, mainly to enhance the signal protection and security of the sensor. Two frameworks enhance the security performance and anti-interference ability of the current signal through supervised learning and frequency authentication of radio frequency signals. To test the signal recognition performance, during the testing phase of sensor devices, a classifier is used to test and classify different sensor signals and devices. The identified data in the sample is defined, as shown in equation (10).

In equation (10), i represents the number of samples to be tested. The equipment probability is recorded, as displayed in equation (11) [22].

In equation (11), p _i represents the probability of sample occurrence. By calculating the information entropy, the data entropy value of the test sample can be obtained, as presented in equation (12).

In equation (12), H represents the entropy value. The other parameters are the same as the above. The total entropy change after increasing the data sample collection is shown in equation (13).

In equation (13), E represents the change in mean entropy. x represents the number of test samples. The other parameters are the same as the above. By analyzing the differences in different samples, the mean entropy of some unknown sensor signals can be obtained, thereby determining the appropriate threshold change for the current signal information and achieving frequency authentication and data encryption for multiple types of wireless passive sensors. Therefore, a wireless passive sensor with machine learning is proposed to enhance the safety and reliability of sensor signals through frequency modulation. Figure 7 shows the testing and training process of frequency data.

Figure 7:

Data testing and training process. Note: Figure 7 illustrates the process of collecting, storing, and training sensor signal data during the analysis and encryption of frequency data. How to input them into the classifier and calculate the mean entropy to determine if they are greater than the current threshold is a key step in evaluating and optimizing sensor signal recognition performance.

In Figure 7, when encrypting frequency data, the signal data of the sensor is first collected. Then, the data information is stored. Unlike the testing signal, the training signal data is added to the training learning. The data signals obtained from the testing and training samples are input into the classifier. The mean entropy is calculated to determine whether the mean entropy is greater than the current threshold. If it is, the current mean entropy is output. If it is not, the iteration of the trained data samples is increased by 1 and the mean is retrained until both are greater than the threshold. Figure 8 shows the frequency authentication and recognition process.

Figure 8:

Frequency authentication and identification.

From Figure 8, in frequency authentication and recognition, firstly, the signal data and signals emitted by the machine are collected and the raw signal data are captured from wireless passive sensors. Next, the collected signals are normalized and denoised. Then, the machine learning method is used to extract features from the signals. Finally, the extracted features are used for signal authentication.

4.1 Result analysis of the anti-interference method

To test the anti-interference ability of wireless passive sensors that currently incorporate frequency authentication and blind equalization algorithm, the input signal of the sensor is modulated using Binary Phase Shift Keying (BPSK) and 16-Quadrature Amplitude Modulation (16QAM). The Signal-to-Noise Ratio (SNR) is 20 dB, the class of the filter is 10, the fading mode of the channel is Reyleigh, and the fading factor is selected as K = 5. The SNR is the ratio of signal power to background noise power, and the higher the SNR, the stronger the anti-interference ability of the model. The error rate represents the ratio of the number of received error bits to the total number of transmitted bits during data transmission. The lower the error rate, the more accurate the data transmission. Throughput refers to the amount of data successfully transmitted by a system per unit of time, with higher throughput indicating higher data transmission efficiency. The SNR is set to 20 dB to simulate a communication environment with EMI from a device. BPSK uses two different phases to represent binary data, which is suitable for scenarios where signal quality requirements are not particularly high. 16QAM is a type of signal that can transmit more data within the same bandwidth, but also has higher requirements for signal quality. The hardware used for the experiment includes a high-performance RFID reader and a customized wireless passive sensor tag. The experimental environment simulates a typical IIoT scenario, which contains a variety of potential sources of interference, such as EMI and multipath effects. The SNR is set to 20 dB. Meanwhile, the output power is set to 500 kW. Two modulation schemes in simulation are used to evaluate the performance of modulation schemes with different complexities and data transmission efficiencies in RFID systems. The step size is 0.0006. The simulated signal runs are 6,000 points. 6,000 signal points are representative samples randomly selected from a larger population, which can reasonably estimate the population characteristics. For statistical testing and model validation, if the effect size is large or the signal in the data is strong, even if the sample size is not large, sufficient statistical power may still be achieved. Meanwhile, small datasets can be used to preliminarily validate the feasibility and potential effectiveness of the designed method, providing a foundation for subsequent large-scale experiments. In the early stages of algorithm development, using small datasets allows for rapid iteration and adjustment, which helps optimize algorithm parameters. The dataset is divided into two signal datasets, each with 3,000 signal points. The simulation results are shown in Figure 9.

Figure 9:

Error rates of readers with different modulation methods for two datasets.

In Figure 9(a) and (b), on dataset 1, the error rate decreases monotonically with the increase of SNR and shows stable convergence. From Figure 9(a) and (c), the BPSK modulation method on dataset 1 has better model convergence, smoother curves, and less curve bending. The error rate decreases with the increase of SNR, showing an overall downward trend. The smooth curve variation indicates good model convergence and stable overall performance. Meanwhile, in Figure 9(c) and (d), on dataset 2, the error rate remains low, indicating excellent performance under complex modulations. Therefore, dataset 2 performs relatively well. To select a better number of readers in the RFID system, a collision rate comparison is performed for different distances and reader quantities, as shown in Figure 10.

Figure 10:

Comparison of the number and number of readers.

In Figure 10, the collision rate represents the conflict changes in the signal reception of the current reader in the system. If the collision rate is high, the overall reception will be weaker. Therefore, the collision rate changed differently for different reader distances under the same number of readers. The collision rate showed an increasing trend with the increase of the reader, and then tended to stabilize. However, the collision rate of readers at different distances varied in their stable state. The collision rate of the reader at 30 m was stable at 0.16. At 32 m, the collision rate at steady state was 0.13. The collision rate of the reader varied at different distances in a stable state, indicating that the distance of the reader has a significant impact on system performance. At 34 m, the reader obtained a collision rate of 0.1. At 36 m, the collision rate when the reader was stable was 0.4. When the reader was approximately 5, the collision rate for almost all distances was in a stable state. From this, the collision rate of the overall system is stable when the reader is selected as 5. This indicates that the overall performance of the system is stable when the number of readers is 5. The experimental simulation process is carried out when the distance is selected as 36 m. To test the impact of adding a new interference algorithm on RFID readers, the signal strength of RFID readers before and after adding the model is compared, as shown in Figure 11.

Figure 11:

Comparison of signal interference before and after method addition. (a) Reading distance before adding algorithm model. (b) Reading distance before adding algorithm model. (c) Reading distance after adding algorithm model. (d) Reading distance after adding algorithm model.

In Figure 11(a) and (b), before the anti-interference method was added, the distance between the reading signal and the received signal of the overall RFID reader/writer was relatively short. The overall reading distance increased with the increase of output power. It first increased and then tended to a stable state. When the output power reached 500 kw, there was also a downward trend in signal reception. The approximate signal reception distance in a stable state was 20 m. Therefore, the overall transmission and reception of sensor signals are at a relatively low level on signal interference, indicating that the distance between signal transmission and reception is limited without interference algorithms. According to Figure 11(c) and (d), after adding anti-interference methods, there was a slight change in the interference distance, but the overall interference distance and trend did not change much. The reading distance had significant changes. Firstly, the curve trend showed an overall upward trend, from the initial receiving reading distance of about 8 to 40. The reading distance continued to increase. Therefore, when using this method to incorporate RFID technology for anti-interference, its anti-interference performance was greatly improved. The signal reception of the sensor was better. This indicates that the proposed method significantly improves the receiving distance and anti-interference performance of sensor signals. This may be attributed to the effective compensation of multipath effects and EMI by the blind equalization algorithm, as well as the enhancement of signal security and interference immunity by the frequency authentication technology. In IIoT, the effective compensation can extend the coverage of the sensor network and reduce data loss. In intelligent transport systems, this enhancement can also optimize traffic signal control and reduce congestion. Due to the interference on the signal, it changed the throughput of the current model. Therefore, when testing the anti-interference ability, the throughput change of the research method is compared with several common methods such as Spatial Filtering (SF), Spread Spectrum Techniques (SST), Adaptive Array Algorithms (AAA), and Waveform Design (WD). The SF and SST are set to explore the data parameter settings using the same method. The SNR is 20 dB, the signal modulation is BPSK, and the interference model is Rayleigh fading. A specific high-pass filter is used. A high throughput rate indicates that the model has stronger anti-interference ability, as shown in Figure 12.

Figure 12:

Comparison of throughput rates of different methods.

In Figure 12, as the exchange volume of signal data increased, the throughput of SF and SST first increased and then showed a downward trend. When the exchange volume was 1, both models showed a downward trend in throughput, and then tended to a relatively stable state. At this time, the throughput of SF was stable at 18 %, and the SST was stable at 22 %. The change of AAA, WD, and the research model were basically consistent. In the initial stage of increasing exchange volume, the throughput showed a rapid growth trend, and then the growth trend relatively decreased. From the change, the throughput value of the research method was higher, at a throughput rate of over 70 %. From this, the proposed method has the best throughput performance, resulting in a better anti-interference effect. The proposed method outperforms other methods on throughput, indicating its stronger anti-interference ability and higher data transmission efficiency. The blind equalization algorithm added to the research method can effectively handle the distortion of the received signal and compensate for multipath effects and EMI, thereby improving signal quality and reducing error rates. Meanwhile, this method improves the security and anti-interference ability of sensor signals through frequency authentication technology, reduces signal conflicts and interference, and enhances the signal transmission reliability. By complementing each other, the anti-interference performance and data transmission efficiency of the system have been jointly improved, achieving higher throughput.

4.2 Anti-interference analysis based on frequency authentication

To test the anti-interference ability and overall effect of the added frequency authentication on sensor signals, the changes in sensor signals are compared. The conventional method is compared with the sensor signal changes when only blind equalization algorithm, frequency authentication, and the combination of the two. The signal band changes are shown in Figure 13.

Figure 13:

Comparison of sensor signal changes after adding different methods.

In Figure 13, a normal sensor signal usually changed in a frequency band after being emitted, but the frequency changed after being disturbed. The overall signal frequency of the image with only frequency authentication was relatively high, but within the frequency range of the sensor. The overall frequency change after only adding the blind equalization model was relatively small compared with when only adding frequency authentication. After adding two methods separately, the frequency change of the signal was the same as that of the interfered sensor signal. After adding two methods, the anti-interference ability of sensor signal transmission was better than that of adding only one method. To compare the differences in anti-interference ability of current methods for wireless passive sensors, the above methods are compared. The results are shown in Table 1. Frequency difference refers to the deviation between the actual frequency and the ideal frequency of a sensor signal after interference. This value reflects the degree to which interference affects the frequency stability of the signal. A small frequency difference indicates that the signal is less affected by interference and the system has stronger anti-interference ability.

In Table 1, the anti-interference ability of the sensor signal varied with different methods. The best anti-interference ability of the research method was 96 %, followed by the model with the blind equalization method, which reached 88 %. Compared with other methods, the anti-interference ability of the research method was 50 % higher than that of SF, 34 % higher than that of SST, 24 % higher than that of AAA, 20 % higher than that of WD, 8 % higher than that of blind equalization model alone, and 22 % higher than that of frequency authentication model alone. The anti-interference ability of the research method is better. The hypothesis testing is conducted. H0: There is no significant difference between the new method and the existing method on this indicator. H1: The new method is superior to the existing method on this indicator. The results are obtained through validation, as shown in Table 2.

From Table 2, there was no statistically significant difference (p < 0.001) in BER, throughput, and signal reception distance among different methods, indicating that H0 was invalid and H1 was valid. This indicates that the new method has better test results.