Wireless sensor signal detection method based on examination and quantification combined design

Abstract

The invention discloses a wireless sensor signal detection method based on review and quantization joint design, which comprises the following steps of: firstly, designing an observation value return method combining a review and quantization mechanism based on a log-likelihood ratio of an observation value, and then designing a maximum likelihood detector based on review quantization for a fusion center; and finally, proposing an examination quantification optimal threshold selection method with the objective of maximizing the performance of the detector. According to the technical scheme provided by the invention, the energy efficiency of the wireless sensor network system is improved. Therefore, the technical scheme provided by the invention can ensure enough signal detection performance while effectively improving the energy utilization efficiency of the sensor.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a wireless sensor signal detection method based on joint design of review and quantification. Background Art

[0002] The Internet of Things is one of the rapidly developing emerging information industries, combining the latest technologies of mobile communications and the Internet. In the application scenarios of the Internet of Things, intelligent perception of the real world is crucial. Wireless sensor networks, with their low cost, flexible deployment and strong self-organizing ability, play a key role in intelligent perception.

[0003] In the perception application of wireless sensor networks, distributed signal detection technology is widely used. This technology is based on a group of wireless sensors and a fusion center to detect specific target signals. These sensors distributed in the detection area use perception components to collect observations from the environment and transmit them to the fusion center after quantification. The fusion center is responsible for receiving the observations returned by all sensors and combining these observations to determine whether the target signal exists.

[0004] Although the fusion center relies on the observations collected by the sensors, the observations are not always valuable because the target signal is not always present. When the target signal does not exist, the collected observations are actually pure environmental noise and interference. In order to avoid excessive consumption of energy and communication resources caused by the transmission of useless observations, the optimization transmission scheme of sensors has been widely studied. The idea is to only transmit valuable observations while ensuring the smallest possible performance loss, so as to reduce the overall energy consumption of the sensor. The existing control schemes are mainly divided into two categories, namely sensor selection schemes and sensor review schemes. In the sensor selection scheme, the fusion center is responsible for determining which sensors will transmit the observation data back; while in the sensor review scheme, the sensor first conducts a preliminary review of the observations, and if it is determined that the observation results have a certain value, they are forwarded to the fusion center. Compared with the sensor selection scheme, the sensor review scheme can determine the transmission status by itself without the need for the fusion center to allocate, so it is more flexible.

[0005] For sensor review schemes, existing technologies often set the observations received by the fusion center as unquantized accurate observations. This means that in actual systems, very high sensor quantization accuracy is required to implement existing review schemes, which runs counter to the requirement for low sensor energy consumption in actual perception systems. Summary of the invention

[0006] In order to solve the limitations and defects of the prior art, the present invention provides a wireless sensor signal detection method based on joint design of review and quantification, comprising:

[0007] M sensors are dispersedly set in a preset area;

[0008] The sensor observes the preset area, obtains the observation value of the target signal source in the preset area, and transmits the observation value to the corresponding review and quantification module;

[0009] The review and quantification module determines whether the observed value is valuable;

[0010] If the review and quantization module determines that the observed value is valuable, the observed value is quantized by single bit to form a quantized value, and the quantized value is transmitted back to the fusion center; if the review and quantization module determines that the observed value is worthless, the sensor remains silent;

[0011] The fusion center determines whether the target signal source exists according to the quantized value and a maximum likelihood detector based on audit quantization.

[0012] Optionally, the step of obtaining the observation value of the target signal source in the preset area

[0013] include:

[0014] The expression of the observation value obtained by the mth sensor is as follows:

[0015]

[0016] in, is the target signal to be detected, It is subject to the mean of 0 and the variance of Gaussian white noise, the observations obey Gaussian distribution:

[0017] Optionally, if the review and quantization module determines that the observation value is valuable, the observation value is single-bit quantized to form a quantized value, and the quantized value is transmitted back to the fusion center; if the review and quantization module determines that the observation value is worthless, the step of the sensor remaining silent includes:

[0018] The log-likelihood ratio of the observed value is calculated, and the expression of the log-likelihood ratio is as follows:

[0019]

[0020] Among them, x m is the observed value, L(x m ) is the log-likelihood ratio of the observed values;

[0021] If the log-likelihood ratio of the observed value is greater than or equal to the review threshold, the observed value is single-bit quantized to form a quantized value, and the quantized value is sent back to the fusion center; if the log-likelihood ratio of the observed value is less than the review threshold, the sensor remains silent.

[0022] Optionally, the step of performing single-bit quantization on the observation value to form a quantized value, and transmitting the quantized value back to the fusion center includes:

[0023] Comparing the observed value with a quantization threshold;

[0024] If the comparison result is that the observed value is less than the quantization threshold, the observed value is quantized to "-1", and "-1" is returned to the fusion center; if the comparison result is that the observed value is greater than or equal to the quantization threshold, the observed value is quantized to "1", and "1" is returned to the fusion center; the expression is as follows:

[0025]

[0026] Among them, η 1，m is the review threshold, η 2，m is the quantization threshold;

[0027] The threshold of the review quantization is calculated according to expression (2) and expression (3), and the expression is as follows:

[0028]

[0029] Among them, τ 1，m is the equivalent review threshold of the mth sensor, τ 2，m is the equivalent quantization threshold value of the mth sensor.

[0030] Optionally, also include:

[0031] According to the Bayesian criterion, the maximum a posteriori probability detector is equivalently converted into a maximum likelihood detector, and the expression is as follows:

[0032]

[0033] Among them, p0 is the prior probability corresponding to H0, p1 is the prior probability corresponding to H1,

[0034] According to the independence of the observation values ​​of each sensor, the expression of the likelihood ratio is transformed into:

[0035]

[0036] Taking the logarithm of expression (6), the log-likelihood ratio obtained using expression (2) is used to represent the detector, and the expression is as follows:

[0037]

[0038] The fusion center expresses the distribution of received observations according to expression (4) as follows:

[0039]

[0040] Substituting expression (8) into expression (2), we can obtain the log-likelihood ratio of the observed value after censoring and quantization, which is as follows:

[0041] L(x m ∈R i )=lnγ 1,j,i / γ 0,j,i ,i∈{-1,0,1}

[0042] Substituting the expression of the log-likelihood ratio of the observed value after censored quantization into expression (7), the expression of the maximum likelihood detector based on censored quantization is obtained as follows:

[0043]

[0044] in, is the detection threshold, T CQ-ML is the test statistic of the detector.

[0045] Optionally, also include:

[0046] The performance of the detector is expressed using the detection probability and false alarm probability, as follows:

[0047]

[0048] The statistical properties of the test statistic are derived using the central limit theorem, as follows:

[0049]

[0050] in, It means that it asymptotically obeys a certain distribution. Indicates that the mean is μ CQ-ML,0 , the variance is Gaussian distribution of

[0051] According to expressions (10) and (11), the detection probability and false alarm probability of the maximum likelihood detector based on audit quantization are obtained, and the expressions are as follows:

[0052]

[0053] Establish an optimization problem with the goal of maximizing detection performance, expressed as follows:

[0054]

[0055] Among them, P c is the average backhaul ratio of the sensor network, expressed as follows:

[0056]

[0057] According to the optimization problem, the optimal value of the review quantization threshold is as follows:

[0058]

[0059] Among them, τ 1，m，opt is the optimal value of the censorship threshold of the mth sensor, τ 2，m，opt is the optimal value of the quantization threshold of the mth sensor.

[0060] The present invention has the following beneficial effects:

[0061] The present invention provides a wireless sensor signal detection method based on the joint design of review and quantization. First, an observation value return method combining review and quantization mechanism is designed based on the log-likelihood ratio of the observation value, and then a maximum likelihood detector based on review and quantization is designed for the fusion center. Finally, a review and quantization optimal threshold selection method with the goal of maximizing the detector performance is proposed. The technical solution provided by the present invention improves the energy efficiency of the wireless sensor network system. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a system model of a wireless sensor signal detection method based on joint design of review and quantification provided in Embodiment 1 of the present invention.

[0063] Figure 2a A schematic diagram of a review and quantization mechanism based on log-likelihood ratio provided in Embodiment 1 of the present invention.

[0064] Figure 2b A schematic diagram of the review quantization mechanism in the observation value domain provided in Example 1 of the present invention.

[0065] Figure 3 This is a flow chart of a wireless sensor signal detection method based on joint design of review and quantification provided in Embodiment 1 of the present invention.

[0066] Figure 4 This is a diagram of simulation results under different signal-to-noise ratios of the wireless sensor signal detection method based on joint design of review and quantization provided in Example 1 of the present invention.

[0067] Figure 5 This is a simulation result diagram of the wireless sensor signal detection method based on joint review and quantization design provided in Example 1 of the present invention under different sensor numbers. DETAILED DESCRIPTION

[0068] In order to enable those skilled in the art to better understand the technical solution of the present invention, the wireless sensor signal detection method based on joint design of review and quantization provided by the present invention is described in detail below with reference to the accompanying drawings.

[0069] Embodiment 1

[0070] 1. Overview

[0071] In order to improve the energy efficiency of wireless sensor network systems, this embodiment proposes a distributed signal detection method based on the joint design of review and single-bit quantization. This embodiment provides a wireless sensor signal detection method based on the joint design of review and quantization, which greatly saves the energy consumption of the sensor by reviewing and quantizing the observation values ​​at the sensor end. In conjunction with the maximum likelihood detector based on review quantization in the fusion center and the optimal value method of the review quantization threshold designed with the goal of optimizing detection performance, the technical solution provided by the present invention can effectively improve the energy utilization efficiency of the sensor while ensuring sufficient signal detection performance.

[0072] 2. Technical route

[0073] This embodiment first designs an observation value feedback method that combines censoring and quantization mechanisms based on the log-likelihood ratio of the observation values, then designs a censoring-quantization maximum likelihood (CQ-ML) detector for the fusion center, and finally proposes a censoring quantization optimal threshold selection method with the goal of maximizing detector performance.

[0074] This section first introduces the system model of the distributed signal detection scenario, and then proposes the review quantization transmission mechanism at the sensor end, the maximum likelihood detector at the fusion center, and the optimal value method of the review quantization threshold.

[0075] 2.1 System Model

[0076] The distributed signal detection system model studied in this embodiment is as follows Figure 1 As shown in the figure. M dispersedly deployed sensors observe the area where the target signal source may exist, and the collected observation values ​​are judged by the review and quantization module. If the observation value is judged to be valuable, it is quantized by single bit and the quantized value is sent back to the fusion center; if it is judged to be worthless, the sensor remains silent. After the fusion center receives all the returned quantized observations, it makes a judgment on whether the target signal source exists based on the CQ-ML detector, completing a detection process.

[0077] In this model, the observation value collected by the mth sensor at a certain moment can be expressed as:

[0078]

[0079] In the formula, is the target signal to be detected, It is subject to the mean of 0 and the variance of Gaussian white noise. In the detection scenario of this embodiment, the distribution parameters of the target signal and the noise are assumed to be known. Therefore, the observed value should also obey the Gaussian distribution: in

[0080] 2.2 Review Quantification Mechanism

[0081] The purpose of reviewing observations is to discard as many worthless observations as possible and return valuable observations. The value of an observation can be measured by its log-likelihood ratio. According to the definition of likelihood ratio, the log-likelihood ratio of an observation can be expressed as:

[0082]

[0083] The likelihood ratio-based censoring mechanism can be summarized as: discard L(x m ) is smaller than the review threshold, and the remaining observations are sent back. Since the observations need to be quantized into single bits before being sent back, the returned observations need to be divided into two parts. The part smaller than the quantization threshold is quantized as "0" and sent back, and the rest is quantized as "1" and sent back. Therefore, the complete review and quantization mechanism on the sensor can be summarized as:

[0084]

[0085] In order to avoid consuming sensor computing power by calculating the log-likelihood ratio, the thresholds of the review quantization are calculated by combining expressions (2) and (3) as follows:

[0086]

[0087] In the formula, τ 1,m ,τ 2,m are the equivalent audit and quantization threshold values ​​of the mth sensor respectively. The optimal value selection method for these two thresholds will be introduced in detail in Section 2.4. Figure 2a and Figure 2b A simplified diagram of this step is given, Figure 2a It represents the three decision domains divided according to the log-likelihood ratio, corresponding to the three states of "keep silent", "return -1" and "return 1" respectively. Figure 2b These are the three decision domains simplified from the log-likelihood ratio formula, corresponding to expression (4).

[0088] 2.3 Maximum Likelihood Detector

[0089] Since the distribution parameters of the target signal are known, the optimal detector is the maximum a posteriori probability detector. Based on the Bayesian criterion, the maximum a posteriori probability detector can be equivalently converted into a maximum likelihood detector:

[0090]

[0091] In the formula, p0 and p1 are the prior probabilities corresponding to H0 and H1 respectively. Taking into account the independence of each sensor observation, the likelihood ratio can be expressed as:

[0092]

[0093] Taking the logarithm of this equation, we can use the log-likelihood ratio obtained from expression (2) to represent the detector:

[0094]

[0095] In the formula, is the detection threshold, T CQ-ML is the test statistic of the detector.

[0096] In order to calculate the above log-likelihood ratio, the distribution of the observations received by the fusion center can be expressed as follows according to expression (4):

[0097]

[0098] Substituting expression (8) into expression (2), we can obtain the log-likelihood ratio expression of the observed value after censoring and quantization: L(x m ∈R i )=lnγ 1,j,i / γ 0,j,i ,i∈{-1,0,1}. Substituting this into expression (7), the complete CQ-ML detector in this scenario can be expressed as:

[0099]

[0100] This embodiment provides a wireless sensor signal detection method based on the joint design of review and quantization, which greatly saves the energy consumption of the sensor by reviewing and quantizing the observation value at the sensor end. In conjunction with the maximum likelihood detector based on review and quantization in the fusion center and the review and quantization threshold optimal value method designed with the goal of optimizing detection performance, the technical solution provided by this embodiment can effectively improve the energy utilization efficiency of the sensor while ensuring sufficient signal detection performance.

[0101] 2.4 Optimal Review Quantization Threshold

[0102] In order to improve the detector performance, this embodiment further sets the optimal review quantization threshold in Section 2.3: 1,m ,τ 2,m , 1≤m≤M, where the detector performance is expressed by the detection probability and false alarm probability as:

[0103]

[0104] Based on the independence of the observations of each sensor, the statistical characteristics of the test statistic can be derived from the central limit theorem:

[0105]

[0106] In the formula, represents asymptotically obeying a certain distribution, The representative mean is μ CQ-ML,0 , the variance is Gaussian distribution. Combining expressions (10) and (11), the detection probability and false alarm probability of the CQ-ML detector are:

[0107]

[0108] Based on this, the optimization problem with the goal of maximizing detection performance can be established as:

[0109]

[0110] Where P c is the average backhaul ratio of the entire sensor network and is defined as follows:

[0111]

[0112] Solving this optimization problem, we can obtain the optimal value of the review quantization threshold as:

[0113]

[0114] Based on the above process, the complete process of the distributed signal detection method proposed in this embodiment based on the review quantization mechanism is as follows: Figure 1 As shown. The whole process can be divided into offline preparation and online working stages. In the offline preparation stage, the fusion center is responsible for calculating the probability distribution of the observation values ​​of each sensor, and calculates the optimal threshold of the review quantization mechanism according to expression (15) and sends it to each sensor. Then the system enters the online working stage, the sensor observes the signal and determines its own feedback status according to expression (4). After receiving all the feedback observation values, the fusion center determines whether the target signal exists according to expression (9).

[0115] 3. Experimental results

[0116] In order to verify the performance of the proposed method, this embodiment simulates the signal detection performance of the following three algorithms under different environments:

[0117] Method 1: The detection method with review and quantification mechanism proposed in this embodiment;

[0118] Method 2: A detection method that only performs single-bit quantization on the observed value without review;

[0119] Method 3: Only the observed values ​​are reviewed, and the returned observed values ​​are not quantified.

[0120] Figure 3 The detection accuracy of the three methods under different signal-to-noise ratios (the ratio of target signal power to noise power) is shown. The number of sensors is set to 100, and the pass rate of method 1 and method 3 is set to P c =0.6. Figure 3 It can be seen that the higher the power of the target signal, the better the detection performance of each method. Under the same target signal power and signal-to-noise ratio conditions, the method proposed in this embodiment is better than method 2. The results show that compared with a pure single-bit quantization mechanism with only two states, the proposed method has three transmission states and can transmit more information to the fusion center. Compared with method 1 and method 3, which also review the observation values, although method 3 does not quantize the observation values, its performance is still inferior to method 1. This is because the method of this embodiment optimizes the CQ-ML detector for the review quantization mechanism, and the performance gain brought by this optimization exceeds the information loss caused by the quantization itself. This result further proves the effectiveness of the CQ-ML detector proposed in this embodiment.

[0121] Figure 4 The detection accuracy of the three methods with different numbers of sensors is shown. The signal-to-noise ratio is set to 0dB, and the review pass rate of method 1 and method 3 is set to P c =0.6. As can be seen from the figure, as the number of sensors increases, the performance of all methods becomes better and eventually converges when the number of sensors is large enough. Compared with Method 2 and Method 3, Method 1 proposed in this embodiment shows obvious performance advantages, which is similar to the simulation results under different signal-to-noise ratios, further proving the effectiveness and advantages of this method.

[0122] 4. Conclusion

[0123] This embodiment studies a distributed signal detection method based on joint design of review and quantization for wireless sensor networks. This method greatly saves the energy consumption of the sensor by reviewing and quantizing the observed values ​​at the sensor end. In conjunction with the CQ-ML detector at the fusion center and the optimal value method of the review and quantization threshold designed with the goal of optimizing detection performance, the method of this embodiment effectively improves the energy utilization efficiency of the sensor while ensuring sufficient signal detection performance.

[0124] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A wireless sensor signal detection method based on joint design of review and quantification, characterized in that: include: M sensors are dispersedly set in a preset area; The sensor observes the preset area, obtains the observation value of the target signal source in the preset area, and transmits the observation value to the corresponding review and quantification module; The review and quantification module determines whether the observed value is valuable; If the review and quantization module determines that the observed value is valuable, the observed value is quantized by single bit to form a quantized value, and the quantized value is transmitted back to the fusion center; if the review and quantization module determines that the observed value is worthless, the sensor remains silent; The fusion center determines whether the target signal source exists according to the quantized value and a maximum likelihood detector based on audit quantization.

2. The wireless sensor signal detection method based on joint design of review and quantification according to claim 1 is characterized in that: The step of obtaining the observation value of the target signal source in the preset area comprises: The expression of the observation value obtained by the mth sensor is as follows: in, is the target signal to be detected, It is subject to the mean of 0 and the variance of Gaussian white noise, the observations obey Gaussian distribution:

3. The wireless sensor signal detection method based on joint design of review and quantification according to claim 2 is characterized in that: If the review and quantization module determines that the observed value is valuable, the observation value is quantized by single bit to form a quantized value, and the quantized value is transmitted back to the fusion center; if the review and quantization module determines that the observed value is worthless, the step of the sensor remaining silent includes: The log-likelihood ratio of the observed value is calculated, and the expression of the log-likelihood ratio is as follows: Among them, x m is the observed value, L(x m ) is the log-likelihood ratio of the observed values; If the log-likelihood ratio of the observed value is greater than or equal to the review threshold, the observed value is single-bit quantized to form a quantized value, and the quantized value is sent back to the fusion center; if the log-likelihood ratio of the observed value is less than the review threshold, the sensor remains silent.

4. The wireless sensor signal detection method based on joint design of review and quantification according to claim 3 is characterized in that: The step of performing single-bit quantization on the observed value to form a quantized value and transmitting the quantized value back to the fusion center comprises: Comparing the observed value with a quantization threshold; If the comparison result is that the observed value is less than the quantization threshold, the observed value is quantized to "-1", and "-1" is returned to the fusion center; if the comparison result is that the observed value is greater than or equal to the quantization threshold, the observed value is quantized to "1", and "1" is returned to the fusion center; the expression is as follows: Among them, η 1，m is the review threshold, η 2，m is the quantization threshold; The threshold of the review quantization is calculated according to expression (2) and expression (3), and the expression is as follows: Among them, τ 1，m is the equivalent review threshold of the mth sensor, τ 2，m is the equivalent quantization threshold value of the mth sensor.

5. The wireless sensor signal detection method based on joint design of review and quantification according to claim 4 is characterized in that: Also includes: According to the Bayesian criterion, the maximum a posteriori probability detector is equivalently converted into a maximum likelihood detector, and the expression is as follows: Among them, p0 is the prior probability corresponding to H0, p1 is the prior probability corresponding to H1, According to the independence of the observation values ​​of each sensor, the expression of the likelihood ratio is transformed into: Taking the logarithm of expression (6), the log-likelihood ratio obtained using expression (2) is used to represent the detector, and the expression is as follows: The fusion center expresses the distribution of received observations according to expression (4) as follows: Substituting expression (8) into expression (2), we can obtain the log-likelihood ratio of the observed value after censoring and quantization, which is as follows: L(x m ∈R i )=lnγ 1,j,i / c 0,j,i ,i∈{-1,0,1} Substituting the expression of the log-likelihood ratio of the observed value after censored quantization into expression (7), the expression of the maximum likelihood detector based on censored quantization is obtained as follows: in, is the detection threshold, T CQ-ML is the test statistic of the detector.

6. The wireless sensor signal detection method based on joint design of review and quantification according to claim 5 is characterized in that: Also includes: The performance of the detector is expressed using the detection probability and false alarm probability, as follows: The statistical properties of the test statistic are derived using the central limit theorem, as follows: in, It means that it asymptotically obeys a certain distribution. Indicates that the mean is μ CQ-ML,0 , the variance is Gaussian distribution of According to expressions (10) and (11), the detection probability and false alarm probability of the maximum likelihood detector based on audit quantization are obtained, and the expressions are as follows: Establish an optimization problem with the goal of maximizing detection performance, expressed as follows: Among them, P c is the average backhaul ratio of the sensor network, expressed as follows: According to the optimization problem, the optimal value of the review quantization threshold is as follows: Among them, τ 1，m，opt is the optimal value of the censorship threshold of the mth sensor, τ 2，m，opt is the optimal value of the quantization threshold of the mth sensor.

Images