JP2016136714A - Radio communication using state measuring method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a radio communication using situation measuring method that can measure the transition rate and occupancy of a channel with high sensitivity and high precision even under an environment having large noise and a large interference effect of electric waves.SOLUTION: In estimation of a channel using situation, the occupancy representing an average use situation, the continuation time of channel use, the transition rate representing a channel switching probability, the average power of a signal emitted from an existing system and the noise power of a self station are estimated at a time, whereby the precision for the channel occupancy and the transition rate can be secured even under an environment having a large noise effect. Each terminal using an application of radio communication can establish communication by performing a measurement according to the measuring method of the present invention and then selecting a channel satisfying QoS requested by the application thereof.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a wireless communication usage status measuring method. Specifically, the present invention relates to a method for measuring and analyzing the usage status of each channel with high accuracy when a wireless terminal performs communication.

  In recent years, modularization of wireless communication systems has progressed, and wireless functions can be obtained at low cost for various devices. As a result, by gathering state information about devices and the environment wirelessly, it becomes possible to find new trends and apply them to device control. These concepts are based on Machine to Machin Communication (M2M), Internet of Things ( IoT), Big Data, and Smart Grid, which are being developed as new business models.

  These wireless communication applications are scheduled to be used in a wide variety of services and applications, and different communication conditions (delay time, transmission speed (throughput), communication quality, etc.) are required. These are called quality requirements (QoS) in wireless communication, and providing a wireless communication environment that satisfies QoS is indispensable for developing a business using wireless communication.

  In wireless communication, it is essential to share a frequency band, which is a wireless communication resource, with a large number of wireless devices in order to enter new applications. These are achieved by dividing the entire frequency band into narrow bands (channels) and sharing the channel in two dimensions of time and frequency. Already, ISM- such as Wi-Fi (registered trademark) and Bluetooth (registered trademark) are used. It is used in a frequency band called BAND.

  Furthermore, the form of the autonomous distributed wireless system in which each terminal selects the channel and timing to be used for each of the divided channels based on its own criteria, and makes it easy for applications that use a wireless communication system to enter the future. Expected as a thing.

  When each terminal selects a channel in an autonomous decentralized wireless system, the channel is observed in advance and used by the local station in order to find a channel that satisfies the required QoS. A method for predicting a possible opportunity has been proposed (Patent Document 1).

  In addition, a probability model focusing on the channel state transition probability, which is generally a Markov model, is applied to a channel usage status whose usage status changes for each application. In the Markov model, the probability of switching the presence / absence of channel operation and the probability of continuous use are collectively evaluated as a transition rate. It has been reported that it is possible to determine whether or not the requested QoS is satisfied by predicting the channel usage based on the Markov model (Non-Patent Document 1). Here, a conceptual diagram of the Markov model is shown in FIG. In the figure, P00, P10, P01, and P11 are transition probabilities indicating the channel usage status. 1 (in use) indicates that the channel is being used, and 2 (not in use) indicates that the channel is not being used. P00 is the probability that the 0 (unused) state will continue, P10 is the probability of transition from 1 (in use) to 0 (unused), P01 is the probability of transition from 0 (unused) to 1 (in use), P11 Means the probability that the 1 (in use) state will continue.

JP 2013-183316 A

Tomoaki Ohtsuki, et al., `` Optimal Channel-Sensing Scheme for Cognitive Radio Systems Based on Fuzzy Q-Learning, "Trans.of IEICE, EB, pp. 283-294, Feb. 2014 Satoshi Takahashi, `` Signal detection characteristics of cognitive radio using differential bandwidth method, '' IEICE Technical Report vol. 109, no. 436, CS2009-111, pp. 221-226, 2010 3 Moon Z-N.Wu, Coding and iterative detection for magnetic recording channels.Kluwer Academic, January 2000.

  Normally, the channel usage varies depending on the application.For example, when downloading data, etc., the channel is used continuously for a long time, but when an application such as streaming is used, the channel can be used in a short time. It tends to be repeated. In this regard, although the method of Patent Document 1 can predict the average channel use opportunity, it cannot measure the waiting time until use or the time that the channel can be used continuously. Especially in applications with high requirements for delay, it was not sufficient.

  Also, channel transition rate and occupancy rate measurement uses a method (carrier sense) that detects radio waves emitted by existing systems and determines the presence or absence of signals. In this method, due to noise in the measurement environment, etc. The reception level fluctuates due to interference (multipath fading) during radio wave propagation. In order to estimate noise power, there is a method using a guard band that is not used by existing systems to avoid adjacent channel interference (Non-Patent Document 2), but this method can be applied only when there is a guard band. It will be limited to the system to do.

  The present invention has been made to solve the above-described problems, and is a wireless communication usage state capable of measuring a channel transition rate and an occupation rate with high sensitivity and high accuracy even in an environment where radio wave interference may occur. An object is to provide a measurement method.

  That is, the present invention includes (1) an energy detection step of obtaining energy values by integrating the power of received signals every predetermined unit time, and using the energy values, the average and variance of signal power of information signals, noise Fluctuation estimation step for obtaining the average and variance of power, and whether the received signal is in a communication signal or noise state, the average and variance of the information signal power obtained in the fluctuation estimation step, and the noise power A likelihood determination step that determines the average and variance using a probability model, and also determines the likelihood, and the update target reception signal to be processed before and after being received in the unit time before and after that A likelihood update step for updating the likelihood based on the state of the received signal and the state transition rate of the received signal excluding the received signal to be updated and the received signals before and after being received in the unit time before and after that, An occupancy rate determining step of determining a state transition rate of the received signal and determining an occupancy rate by counting the state transition pattern of the state of the received signal updated in the writing likelihood update step, This is a communication usage measurement method.

  In another aspect of the present invention, (2) the received signal to be updated in the likelihood update step is a signal having a likelihood closest to 0.5 among the received signals. It is a wireless communication usage status measurement method described.

  Another aspect of the present invention is the wireless communication usage status measuring method according to claim 1 or 2, wherein (3) the likelihood update step is repeated twice or more.

  Further, another aspect of the present invention further includes: (4) a feedback step of performing the fluctuation estimation step again using the occupation rate determined in the occupation rate determination step as an input value. 4. The wireless communication usage status measuring method according to any one of items 3 to 3.

  The wireless communication usage status measurement method according to the present invention is based on the estimation of the channel usage status, the occupation rate indicating the average usage status, the transition rate indicating the channel usage duration and switching probability, and the signal generated by the existing system. This is a technique for ensuring the accuracy of channel occupancy and transition rate estimation even in an environment where the influence of noise is large by estimating the average power and the noise power of the local station at once. Each terminal that uses a wireless communication application can perform a measurement using the measurement method according to the present invention, and can then select a channel that satisfies the QoS required by the application and transmit a communication request.

  The wireless communication usage status measuring method according to the present invention can be implemented by a computer system in which a program is recorded or a program is read from another recording medium. The computer system here includes hardware such as a mobile phone, a personal computer, or a peripheral device thereof. 2 and 3 in the present application are conceptual diagrams, and each physical configuration and processing step illustrated can be configured by hardware, software, and a combination thereof.

It is a conceptual diagram of a Markov model. It is a flowchart which shows the flow of a process of this invention. It is a conceptual diagram of the system environment which this invention makes object. It is an example of a histogram analysis. It is an image when the probability is around 0.5. It is a procedure of likelihood update step. It is a graph of a simulation result about the present invention. It is a graph of the mean square error of P11. It is a graph of the mean square error of P00. It is a graph of the mean square error of occupation rate P1. It is a graph of the correct answer rate of the estimation in a present Example.

  Below, the form for implementing the radio | wireless communication utilization condition measuring method which concerns on this invention is demonstrated.

  A processing flow of the measurement method according to the present invention is shown in FIG. The present invention is mainly constituted by five processes. In the first process (10), energy is detected. In the second process (20), fluctuations in detected energy and noise are estimated. Specifically, the average and variance of information signal power and the average and variance of noise power are estimated. The third process (30) evaluates the energy reliability (likelihood). In the fourth processes (40) and (50), the reliability evaluation value (likelihood evaluation) obtained by the previous process is updated by collecting external information. The last process (60) estimates the occupation rate and the transition rate.

  A conceptual diagram of the system environment targeted by the present invention is shown in FIG. In the system environment shown in the figure, there is one existing system, and that system issues an information signal at its own discretion. In the system environment, there is a sensor that detects whether the channel of the existing system is used. The sensor pulls in the signal of the existing system from the antenna and observes the channel usage status.

<Step of detecting energy>
In the energy detection step (10), as shown in FIG. 3, the received signal, which is an analog signal detected from the antenna, is sampled at a constant interval, and a power sample is obtained by squaring the voltage sample. Then, a certain number of samples are summed to obtain an energy sample. In the present specification, a time series of energy samples is referred to as an energy series. Noise generally follows a Gaussian process with mean 0 variance s _n ² , which can be expressed as N (0, s _n ² ). When the information emitted from the information source is modulated, the modulated signal becomes a Gaussian process according to N (0, s _s ² ) by making the information an irregular process. Here, s _s ² is the average power of the signal.

  The received signal detected by the antenna is a communication signal in which an information signal and noise are added if the existing system uses a channel, and only the noise if the existing system does not use the channel. . Here, noise refers to an irregular signal mainly generated in the receiver circuit, and means a signal detected even though the existing system does not transmit an information signal using the observation target channel. To do. Therefore, the noise includes a case where it is negligibly small compared to the information signal.

When the sensor detects a signal and the existing system uses a channel, the energy value obtained by carrier sense is an energy sample obtained from a communication signal obtained by adding an information signal and noise. On the other hand, when the existing system does not use the channel, the energy value is only noise. As a result of the processing by the energy detection step, the energy value of noise can be modeled by N (Ks ² _n , Ks ⁴ _n ). Here, K represents the number of samples when the power samples are added. The energy sample of the communication signal in which the information signal and the noise are added is N (K (s ² _s + s ² _n ), K (s ² _s + s ² _n ) ² ).

<Step of estimating signal fluctuation>
Since the energy sequence obtained by the energy detection step does not specify the channel usage of the existing system, the probability of the energy sample of the communication signal that each energy is a stochastic process of the noise energy sample or the noise plus the information signal It is not possible to specify which of the processes will be performed. Therefore, the energy sample sequence can be considered as a set of energy samples in which two different stochastic processes are mixed. In the signal fluctuation detection step (20) from such an energy sample set in which a plurality of stochastic processes are mixed, the average value of noise power, the variance value, the average value of signal power of the communication signal, and the variance value are obtained by the following procedure. presume.

<Signal fluctuation estimation method>
(1) Limit the energy sequence to a finite time. Here, M is the number of energy samples existing within a finite time.
(2) The range from the minimum value to the maximum value of the energy sequence is equally divided into L, and a discrete sequence composed of the values of the respective division points is defined as y ′.
(3) Histogram analysis of energy series. An example of histogram analysis is shown in FIG. The energy range searched by the histogram is y ′. After the analysis, the frequency of each energy is normalized by the total number to calculate the probability of occurrence of each energy. Let p ₁ be the sequence of occurrence probabilities. Here, p ₁ is a horizontal vector.
(4) s ² _s and s ² _n are set to values obtained by equally dividing the range from the maximum value of the energy sequence to the minimum value by L. That is, when the division interval is Dy, the average power s ² _s of the information signal and the average power s ² _n of the noise are as follows.
Here, k and l are integer values from 0 to L-1, respectively, and y _min is the minimum value of the observed energy value.
(5) The combined probability density function γ ₁ N (Ks ² _n , Ks ⁴ _n ) + γ ₂ N (K (s ² _s + s ² ) is obtained from s ² _s and s ² _n obtained as k = 0, l = 0. _n ), K (s ² _s + s ² _n ) ² ) are determined, and y ′ is substituted as a discrete sequence. Here, γ ₁ and γ ₂ correspond to the probability that the signal of the existing system does not exist and the probability that it exists (the probability that the signal of the existing system exists is equal to the occupation ratio), respectively, and γ ₂ = 1−γ ₁ . Each of γ _{1 and} γ ₂ is given an initial value of 0.5, and can be updated to the value of the occupation rate when the occupation rate is obtained in a later step. By performing processing again using the updated value, it becomes possible to perform measurement with higher accuracy. The occurrence probability for each element of the discrete sequence y ′ is calculated by multiplying the quantization density Dy from the probability density obtained in this way. Let p _{2 be} the occurrence probability. Here p ₂ is a lateral vector.
(6) The probability squared error DP is calculated as follows:
Here, (·) ′ is an operator indicating transposition.
(7) DP is calculated for all possible values of k and l, and the values of k and l when DP is the smallest are assumed to be estimated values s ² _{n and} s ² _s .

<Likelihood evaluation step>
As a result of the process of estimating the fluctuation of the signal, an average value of the information signal power and the average value of the estimated information signal power and noise power are obtained. In the next step (30), the likelihood indicating the possibility of being a communication signal and noise is calculated as follows using the fact that each sample follows one of two probability processes. First, the probability density function (N (K (s)) assuming that the nth (n = 0, 1,…, M-1) energy sample follows the stochastic process of the communication signal with noise and information signal added. ² _s + s ² _n ), K (s ² _s + s ² _n ) ² )), the probability density function (N (Ks ² _n , Ks ⁴ _n )) The logarithmic function with the natural logarithm as the base is applied. As a result, the log likelihood function l _n in the nth sample is given by the following equation.
Thus, the probability q _n that the n-th energy sample follows the stochastic process of the communication signal in which the information signal and the noise are added is given by the following equation ₍ 4) or (5).
Here, Equation 5 is an expression obtained based on a paper, and the probability can be obtained effectively when the likelihood evaluation is accurate (Non-patent Document 3). When the equation 4 is used when the probability q _n is obtained, the absolute value of the probability is lower than that obtained using the equation 5. However, in the equations 4 and 5, the tendency of the determination result is similar.

<Step of updating likelihood>
As a result of the process of evaluating the likelihood, the probability that each energy sample is a communication signal obtained by adding an information signal and noise is obtained (step (40)). FIG. 5 is a graph showing the relationship between the probability of being a communication signal and the probability of noise for an energy sample.
According to FIG. 5, when the detected energy value is larger than a certain value, there is a high probability that the information signal and noise are added, and conversely, when the detected energy value is smaller than a certain value, only the noise is present. It can be evaluated that the probability is high. Therefore, depending on the detected energy value, an ambiguous energy sample may be detected whether it is a communication signal or noise. When such an energy sample is used to determine whether or not there is access, the probability of false alarms and false detections is high.
Therefore, next, as a likelihood evaluation update step (50), the probability of determining the communication signal as an information signal and noise added is updated, and processing for further increasing the accuracy is performed. The procedure of the likelihood update step is shown in FIG.

(1) One of the energy samples with the probability q _n (n = 0, 1, ..., M-1) that is a communication signal with information signal and noise added, the one closest to the value of 0.5 Select this energy sample as the lowest likelihood sample. If there are a plurality of corresponding samples, for example, the oldest time sample is selected. Here, the time of the sample with the lowest likelihood is k (k = 0, 1,... M−1).

(2) Except for the lowest likelihood sample, if the probability q _n of each energy sample exceeds the threshold value 0.5, label 1 as the energy sample of the communication signal to which the information signal and noise are added. If it falls below the threshold, label it as a noise energy sample.

(3) Count the number of occurrences of transition patterns (4 types, 1 to 0, 0 to 1, 1 to 1, 0 to 0) for each energy value label. Here, the frequency numbers of the respective transitions are u ₁₀ , u ₀₁ , u ₁₁ , u ₀₀ , respectively. At that time, the transition patterns before and after the lowest likelihood sample are not included in the number.

(4) Update the probability r′k that the energy sample of the communication signal in which the information signal and noise are added in the lowest likelihood sample is obtained as follows.
However, when k = 0, {(qk-1P11) + ((1-qk-1) P01)} = 1. On the contrary, when k = M−1, {(q _{k + 1} P ₁₁ ) + ((1-q _{k + 1} ) P ₁₀ )} = 1.

(5) Next, a probability t′k at which a noise energy sample is given is calculated by the following equation.
However, when k = 0, it is assumed that {(q _k-1 p ₁₀ ) + ((1-q _k-1 ) p ₀₀ )} = 1. Conversely, when k = M−1, {(q _{k + 1} p ₀₁ ) + ((1-q _{k + 1} ) p ₀₀ )} = 1.

(6) The probability q ′ _k that an energy sample of a communication signal in which an information signal and noise are added is obtained by normalization processing is given by the following equation.
From the result obtained in Equation 8, the probability that an energy sample of the communication signal is obtained is updated as q _k = q ′ _k . In the wireless system according to the Markov model, this update process has dependency on adjacent states, and likelihood can be updated by collecting likelihood information obtained from the dependency.

(7) Steps 1 to 6 are repeated, and the process ends when the probability q _n of all received signals in the energy sequence is more than a certain distance from 0.5, or when the repetition is repeated a certain number of times.

<Transition rate, occupation rate measurement step>
In the transition rate / occupancy rate measurement step (60), when the probability q _n which is a communication signal obtained by adding the information signal and noise calculated from each energy sample value exceeds 0.5 or more, it is determined as 1 and when it is below 0.5 And the transition rate is estimated using Equation (9). The frequency number of 1 and the frequency number of 0 are calculated, and the occupation rate is obtained by calculating the frequency number of P1 = 1 / (frequency number of 1 + frequency number of 0).

<Return step>
As a result of obtaining the occupation rate P ₁ by the processing of the transition rate and occupation rate measurement step, this can be applied as γ ₁ = 1−P ₁ , γ ₂ = P ₁ , and the fluctuation estimation process can be performed again. This can be applied when further improvement in occupation rate can be expected.

In order to show the effectiveness of the measurement method according to the present invention, numerical evaluation by computer simulation was performed. The number of accumulated samples K of the integrator in the energy detection step is set to 30, and the number of sections for searching in the fluctuation estimation step is set to L = 100. In addition, the number of energy sample series M was set to 2000. The access model of the existing system is assumed to follow the Markov model shown in FIG. 1, and the measurement accuracy characteristics of the transition rates (P ₁₁ , P ₀₀ ) when P ₁₁ = 0.8 and P ₀₀ = 0.7 are estimated. In this embodiment, the measurement occupation ratio is not fed back from (60) to (20) in the processing procedure of FIG. 2, and the coefficient γ1 = γ2 = 0.5 in (20) is constant.

Here, MSE is an abbreviation of Mean Square Error and is given by the following equation.
In the formula, a is an identifier indicating 00, 11 or 1 in the case of the occupation ratio. Pa (-) represents an estimated value, and Pa represents _a true estimated value. Thus, the lower the MSE value, the better the measurement accuracy.

(A), (b), and (c) of FIG. 7 show the measurement accuracy of p ₀₀ , p ₁₁ , and p ₁ , respectively. In the figure, as a conventional method, the processing up to before the likelihood update is applied, the probability q _n of the received signal of each symbol is determined as 0.5 as a threshold value, and the results when the transition rate and occupancy rate are measured are shown. ○ indicates. The measurement results according to the present invention are indicated by ●. From the figure, it can be seen that the measurement method according to the present invention achieves 5% MSE even in an environment where the signal power to noise power ratio (SNR) is about 1 dB lower.

In the present invention, numerical evaluation was performed by computer simulation when the probability q _n according to the stochastic process of the communication signal in which the information signal and noise were added for the nth energy sample was obtained using Equation 5. The number of accumulated samples K in the integrator in the energy detection step was set to 2000, and the number of sections for search in the fluctuation estimation step was set to L = 100. Further, the number of energy sample series M was set to 30. The access model of the existing system is assumed to follow the Markov model shown in Fig. 1, and (P ₁₁ , P ₀₀ ) = (0.8, 0.7) and (P ₁₁ , P ₀₀ ) = (0.7, 0.6) are assumed. The transition rate and occupancy rate were estimated.

8 to 11 show the mean square errors and correct answer rates of P ₁₁ , P ₀₀ , and P ₁ with respect to SNR by changing the threshold T in the likelihood evaluation in this example. 8 and 9, it can be seen that the measurement errors of P ₁₁ and P ₀₀ have improved characteristics compared to before the likelihood update process regardless of the set values P ₁₁ = 0.8 and P ₁₁ = 0.6. I can confirm. From Fig. 10, the improvement effect of the estimation accuracy of P1 was recognized when P11 = 0.8. Moreover, although the estimation accuracy of P1 has deteriorated even at P11 = 0.6, MSE = 0.05 or less has been achieved, and it can be said that the amount of deterioration is small.

In the transition rates p ₀₀ , p ₁₁ , and the occupation rate p ₁ shown in FIGS. 8, 9, and 10, measurement errors due to the presence or absence of the likelihood update step were compared. In the figure, the data plotted as C shows the case where there is no likelihood update step, and T = 0.01 to T = 0.03 shows the case where the likelihood update is performed with the threshold value being 0.01 to 0.03, respectively. . From the figure, it can be seen that by performing the likelihood update step, a low MSE is achieved even in an environment with a low SNR.

As described above, the measurement method according to the present invention occurs in wireless communication by collecting likelihood information from neighboring samples rather than determining the usage status of wireless communication based on the energy of each sample. Eliminates power fluctuations and achieves high-precision judgment. As a result, the measurement with high accuracy and sensitivity is achieved even with lower signal power for both the occupation ratio and the transition ratio.
According to the present invention, it is possible to estimate a channel that satisfies the QoS required by an application and to realize a system that performs efficient communication.

DESCRIPTION OF SYMBOLS 10 Energy detection step 20 Fluctuation estimation step 30 Likelihood determination step 40 Transition rate temporary determination step 50 Likelihood update step 60 Occupancy rate determination step

Claims (4)

An energy detection step of integrating the power of the received signal every predetermined unit time to obtain an energy value;
Fluctuation estimation step for obtaining the average and variance of the signal power of the information signal and the average and variance of the noise power using the energy value;
Whether the received signal is in a communication signal or noise state is determined using the average and variance of the information signal power acquired in the fluctuation estimation step, the average and variance of the noise power, and a probability model. A likelihood determining step for determining the likelihood by
For the update target received signal to be processed, the status of the received signal before and after received in the unit time before and after that, and the update target received signal and the received signal before and after received in the unit time before and after that are excluded. A likelihood update step for updating the likelihood based on the state transition rate of the received signal;
Counting the state transition pattern of the state of the received signal updated in the likelihood update step, determining the state transition rate of the received signal, and the occupation rate determining step,
A wireless communication usage status measuring method comprising:   2. The wireless communication usage status measuring method according to claim 1, wherein the received signal to be updated in the likelihood update step is a signal having a likelihood closest to 0.5 among the received signals.   The wireless communication usage status measurement method according to claim 1 or 2, wherein the likelihood update step is repeated two or more times.   The wireless communication utilization method according to any one of claims 1 to 3, further comprising: a feedback step of performing a fluctuation estimation step again using the occupation rate determined in the occupation rate determination step as an input value. Situation measurement method.