JP5252430B2 - Signal detection method, program, information storage medium, and sensor - Google Patents

Description

  The present invention relates to a signal detection method for detecting an analog signal for wireless communication.

  Cognitive radio is a system that finds and uses the time and frequency gaps of a licensed system. For this reason, it is necessary to sufficiently perform carrier sense before transmitting information. Japanese Patent Laying-Open No. 2008-53830 (Patent Document 1) discloses a cognitive radio communication device that maintains sufficient carrier sense performance and has high transmission efficiency while keeping the hardware scale to a minimum. Has been. This wireless communication apparatus includes a cognitive engine for performing cognitive wireless communication.

  By the way, in order to perform carrier sense, a cognitive engine needs to detect a primary signal (spectrum sensing) in the presence of noise in the background. Therefore, a detector for detecting the energy of the primary signal is known.

  FIG. 7 is a block diagram showing a configuration of a conventional detector. The conventional detector is configured as shown in FIG. 7 and includes function blocks such as a squaring device, an integrator, and a comparator. When a signal is input to the detector, the signal output is squared by a squaring device, and then the squared signal output is compared with a square obtained by squaring a predetermined threshold value by a comparator. The Here, the {threshold / threshold} threshold is determined by the power of the environmental noise caused by the failure of the receiver itself and the surroundings. Then, by comparing, it is possible to detect a signal and obtain a sensing result even in the presence of noise.

However, since the sensing result reflects the situation at a specific place and time, the sensing output also changes with the passage of time. For this reason, the SNR (signal to noise ratio) input to the detector may be small. In this case, in the above detector, the detection performance (particularly the detection accuracy) deteriorates unless a large number of input signal samples are prepared. On the other hand, increasing the detection performance of the detector requires more complicated calculations, and thus the number of function blocks increases.
JP 2008-53830 A

  The main object of the present invention is to provide a signal detection method and a sensor capable of enhancing the detection performance of an analog wireless signal used for wireless communication. It is another object of the present invention to provide a program and an information storage medium for causing a computer to execute the method.

  In addition, the present invention provides a signal detection method and sensor that can prevent the calculation complexity from becoming excessively high even if the detection performance is improved when detecting an analog signal used for wireless communication. The other purpose is to provide. It is another object of the present invention to provide a program and an information storage medium for causing a computer to execute the method.

  The present invention basically relates to a wireless signal detection method for detecting an analog wireless signal for wireless communication. In this radio signal detection method, the detector 10 is used when detecting an analog radio signal.

  The detector 10 includes a buffer 13 for storing a digital signal corresponding to an energy value of an input signal input from the outside, an autocorrelation device 14 for performing autocorrelation processing on the signal, and a fast Fourier transform for the signal. An FFT device 16 that performs processing, a comparator that compares a predetermined upper limit value with respect to an energy value of noise in the background of the analog wireless signal, and an energy value of the signal, and the comparator , And a determination device 18 for determining whether or not the analog radio signal to be detected has been detected. Note that the upper limit value of noise may be changed by computer control. In this case, it is preferable to change the upper limit value in accordance with a change in the noise output value (for example, in proportion).

  The method includes an accumulation step (S14), an autocorrelation step (S16), an averaging step (S18), an FFT processing step (S20), a comparison step (S22), a determination step (S24), Have

  In the accumulation step (S14), digital signals are accumulated in a predetermined number of samples in the buffer 13 of the detector 10. In the autocorrelation step (S16), the autocorrelation device 14 reads out a digital signal of a predetermined number of samples stored in the buffer 13, and performs autocorrelation processing. Thereby, sine wave detection becomes possible. This effectively reduces the problems that occur during detection. As a result, the detection performance of the detector 10 can be improved. In the averaging step (S18), the autocorrelation step (S16) is repeated. As a result, a periodic signal is obtained from the signal after autocorrelation processing obtained in each of the autocorrelation steps (S16). Since the averaging step (S18) is performed after the autocorrelation step (S16), the complexity can be minimized. For this reason, the required computational complexity can be prevented from becoming excessively high. Further, by minimizing the complexity, the power consumption of the device in which the detector 10 is mounted can be suppressed.

  In the above method, in the FFT processing step (S20), the FFT device 16 performs fast Fourier transform processing on the periodic signal. As a result, a spectrum is obtained. In the comparison step (S22), the comparator compares the spectrum obtained in the FFT processing step (S20) with the upper limit value. In the determination step (S24), the determination device 18 determines whether or not the analog radio signal to be detected can be detected from the input signal using the comparison result by the comparator. Through the steps described above, information on the analog radio signal to be detected is acquired from the input signal.

  In a preferred embodiment of the signal detection method of the present invention, the autocorrelation device 14 is a composite device having a plurality of shift registers, a plurality of multipliers, and a plurality of adders. In the averaging step (S18), a composite device is used. The device performs autocorrelation processing on the delay time, and obtains a periodic signal by averaging the obtained signals after the autocorrelation processing over the observation period.

  Another preferred aspect of signal detection according to the present invention is a program for causing a computer to execute the steps of the above-described wireless detection method. Alternatively, a computer-readable information storage medium storing a program for causing a computer to execute the steps of the wireless detection method. Even if it is an aspect like these, the effect mentioned above can be produced.

  Still another preferred embodiment of the signal detection of the present invention is a sensor. This sensor has the detector 10 and an input unit for inputting a signal to the detector 10. Even if it is such an aspect, the effect mentioned above can be produced.

  ADVANTAGE OF THE INVENTION According to this invention, the signal detection method and sensor which can improve the detection performance of the analog radio signal used for radio | wireless communication can be provided. In addition, according to the present invention, it is possible to provide a signal detection method and a sensor that can reduce the required computational complexity even if the detection performance is improved when detecting an analog signal used for wireless communication. it can. Further, it is possible to provide a program and an information storage medium for causing a computer to execute the above method.

  Hereinafter, the best mode for carrying out the present invention will be described with reference to the drawings. However, the form described below is an example, and can be appropriately modified within a range obvious to those skilled in the art.

  FIG. 1 is a block diagram schematically showing the configuration of the detector 10 of the present invention. The detector 10 of the present invention is mounted on a wireless communication terminal (for example, a mobile phone) and is connected to an input unit provided in the wireless communication terminal. The input unit of the wireless communication terminal is for receiving an analog wireless signal in the wireless communication area, and the analog wireless signal received by the input unit is input as an input signal into the wireless communication terminal. Therefore, it can be said that a sensor for wireless communication is configured by the input unit and the detector 10. The wireless communication terminal is preferably a terminal capable of performing cognitive wireless communication. This is because the detection performance of a primary signal, which is an example of an analog radio signal, is improved by mounting the detector 10 of the present invention on a cognitive radio communication terminal (described later).

  The detector 10 is for detecting an analog wireless signal for wireless communication. As shown in FIG. 1, an analog-digital (A / D) converter 11, an averaging processing unit 12, and a high-speed A Fourier transform (FFT) device 16 and a determination device 18 are included. The averaging processing unit 12 includes a buffer 13 and an autocorrelation device 14. Each unit of the detector 10 is preferably configured by software such as a function block because the apparatus can be downsized. However, at least a part of the detector 10 may be configured by hardware or a combination thereof.

  The A / D converter 11 converts an energy value of an input signal input from the outside into a digital signal. Specifically, the input signal input from the input unit of the wireless communication terminal is converted into a digital signal, and the digital signal is input to the buffer 13. By converting to a digital signal, the analog signal is converted into a collection (stream) of sample signals.

  The averaging processing unit 12 is a functional block that repeatedly performs predetermined processing in accordance with the number of digital signal samples input from the A / D converter 11. The buffer 13 stores data such as digital signals. In practice, the buffer 13 stores the digital signal input from the A / D converter 11 by a predetermined number of samples (for example, N). As a result, the number of samples matching the target detection performance is prepared.

  The autocorrelation device 14 is for performing autocorrelation processing on the signal. In this embodiment, the autocorrelation device 14 is a composite device having a plurality of shift registers, a plurality of multipliers, and a plurality of adders. A known device (for example, a multiplexer) can be used as the composite device. Therefore, the autocorrelation process is automatically performed.

  Specifically, the autocorrelation device 14 reads out the digital signal stored in the buffer 13 every predetermined number of samples (for example, N), and performs autocorrelation processing on the read digital signal, thereby The signal after correlation processing is acquired. The signal after autocorrelation processing is also converted into a collection (stream) of sample signals. This autocorrelation process is repeatedly executed in the averaging processing unit 12. Accordingly, the autocorrelation device 14 acquires a periodic signal (smoothed pseudoperiodic signal) from a plurality of signals after autocorrelation processing. The obtained periodic signal is input to the FFT device 16. In the averaging processing unit 12, it is preferable that the autocorrelation process is performed with respect to the delay time, and the obtained signal after the autocorrelation process is averaged over the observation period. For example, when autocorrelation processing is performed on N samples with respect to a delay time of 0 to L, 2 × L + 1 samples are output. Here, the maximum value L is selected so as to satisfy (L / R) <1 / B. Here, R is a sampling rate by the A / D converter 11, and B is a bandwidth.

  The FFT device 16 is for performing a fast Fourier transform process on the signal. Specifically, the FFT device 16 performs a fast Fourier transform process on the periodic signal input from the autocorrelation device 14. Thereby, a spectrum (frequency domain) is obtained. Here, it is preferable to efficiently determine the detection position in the frequency spectrum by grouping bins (frequency bins) related to the frequency of the FFT output into groups composed of a plurality of subbands.

  The decision device 18 has a comparator. Then, the determination device 18 is for determining whether or not the analog wireless signal to be detected can be detected from the input signal using the comparison result by the comparator. Therefore, the output signal of the determination device 18 is 1 when the detection target analog radio signal can be detected (when detection is successful), and the detection target analog radio signal cannot be detected (detection). 0) when there is no input signal.

  The comparator of the decision device 18 is for comparing the threshold value with the energy value of the signal. Here, the threshold value is stored in advance in the detector 10 and is an upper limit value predetermined for the energy value of noise in the background of the analog wireless signal. Therefore, noise that exceeds the upper limit value cannot be detected by the detector 10. Specifically, the comparator compares the upper limit value with the spectrum input from the FFT device 16. Here, when the frequency bins of the FFT output are grouped, the average value of each subband is compared with the threshold value. Note that the upper limit value of noise may be changed by computer control. In this case, it is preferable to change the upper limit value in accordance with a change in the noise output value (for example, in proportion).

  Then, by configuring the detector 10 as described above, the detector 10 or the wireless communication terminal including the detector 10 can receive information (for example, a signal related to an analog wireless signal to be detected) from the input signal received by the input unit. Intensity and its duration) can be obtained.

  Next, an operation of the wireless communication terminal including the above-described detector 10 will be described. The operation of the detector 10 corresponds to executing the steps of the signal detection method of the present invention. FIG. 2 is a flowchart showing an operation procedure of the detector 10. In FIG. 2, S indicates a step. The program corresponding to the flowchart corresponding to FIG. 2 is stored in the information storage medium. Then, the program corresponding to each step in FIG. 2 is performed by reading and executing the program by the computer built in the information processing terminal or the detector 10.

  In FIG. 2, first, the wireless communication terminal receives an analog wireless signal in the wireless communication area at the input unit (S10). The received analog wireless signal is input to the A / D converter 11. The A / D converter 11 converts the input analog radio signal into a digital signal (S12), inputs the obtained digital signal into the buffer 13 by a predetermined number of samples, and thereby accumulates it in the buffer 13 ( S14).

  Subsequently, the autocorrelation device 14 reads the digital signal stored in the buffer 13 by a predetermined number of samples and performs autocorrelation processing (S16). Thereby, a signal after autocorrelation processing is obtained. And the averaging process part 12 repeats the process of step S14-S16 (S18). Thereby, a periodic signal is obtained from the signal after autocorrelation processing obtained in each of the autocorrelation steps (S16). This periodic signal is input to the FFT device 16.

  Then, the FFT device 16 performs a fast Fourier transform process on the periodic signal to obtain a spectrum (S20). The obtained spectrum is compared with an upper limit value which is a threshold value by a comparator (S22). Then, the comparison result by the comparator is input to the determination device 18. The determination device 18 uses the comparison result input from the comparator to determine whether or not the analog radio signal to be detected can be detected from the input signal (S24). Thereafter, if necessary, an analog radio signal that has been successfully detected is analyzed, and information about the analog radio signal (for example, signal strength and its duration) is acquired.

  According to the processing (signal detection method) of FIG. 2, the detection performance (that is, the reliability of sensing) of an analog wireless signal used for wireless communication can be improved as compared with the conventional detector 10. Furthermore, even if the detection performance is improved in this way, the required calculation is about the FFT processing, so that the complexity does not become excessively high. As a result, power consumption can be reduced.

  Moreover, according to the aspect mentioned above, the sensor suitable for the radio | wireless communication terminal which performs each step of the signal detection method of FIG. 2 and the said radio | wireless communication terminal is provided. In particular, in a cellular phone or the like, since the maximum power and calculation processing capacity are limited, it is very useful to mount the sensor (detector 10) of the present invention.

  Next, examples of the present invention will be described. In particular, the performance of the detector 10 of the present invention and the performance of a conventional detector will be described in a comparable manner.

  In this embodiment, a simulation was performed using a wireless communication system having a primary system and a secondary system as a model as a wireless communication system. In the simulation, the following settings were made. First, in this model, it is assumed that the primary user has occupied a part of a predetermined radio communication (RF) spectrum for a while. Second, the secondary system is capable of cognitive radio communication, and when sensing the RF environment, the frequency band is not occupied by the primary system and the spectrum is used for convenience. Third, the secondary receiver side cannot obtain preliminary knowledge (information) about the primary signal. In this embodiment, an AWGN (additive white Gaussian) channel is considered as a simple scenario.

When the received signal is expressed in the discrete time domain under the above setting conditions, the following equation is obtained. In the following equation (1), s [n] is a sample of the primary signal, and w [n] is a sample of additive white Gaussian noise.
r [n] = s [n] + w [n] (1)

Then, in order to perform the spectrum estimation, considering the sample of the sample number of inputs N _S as a recording length performed once. Actually, s [n] can be composed of two or more communication signals that occupy different parts of the RF spectrum, and moreover, multiple primary system signals can exist around the cognitive radio communication (CR). I made it. Furthermore, r [n] was assumed to be stationary during the sensing period.

  Then, in order to examine how to improve the detection performance of the cognitive radio communication terminal, a scheme for detecting signal energy in three cases (Reference Examples 1 and 2 and Example 1) was examined by simulation. Reference Example 1 is a simulation when a conventional detector as shown in FIG. 7 is used, Reference Example 2 is a simulation when a periodic stationarity detector is used, and Example 1 is shown in FIG. This is a simulation when the detector 10 according to the present invention as shown in FIG. 1 is used (that is, when autocorrelation is used).

  First, Reference Example 1 will be described.

  In Reference Example 1, a simulation was performed using a conventional detector as shown in FIG. In this simulation, it was assumed that the RF environment was stationary. However, in practice, since the RF environment is not stationary, it is needless to say that it is preferable to appropriately estimate the threshold value as time passes.

  In this simulation, as shown in FIG. 3, energy detection is performed in the frequency domain. The main reason is that if a signal is detected, the position of the frequency is known. Note that energy detection can also be performed in the time domain, but in this case, a plurality of filters are required to implement the receiver. As these filters, filters in which the center frequency of the filter matches the center frequency of the signal assumed by the receiver are used.

  When operating the receiver, basically, the PSD (power spectrum density) of the received RF signal is first calculated, and the result is averaged over a plurality of observation times. Considering that the number of samples for calculating the PSD is limited, it is possible to first obtain an unbiased estimate by multiplying the sample sequence by an appropriate window function. However, the window function was not considered at all in this simulation.

And regarding the energy detection, the decision statistic was calculated according to the following equation (2). In addition, when Expression (1) is substituted into Expression (2), the following Expression (3) is obtained.
z [n] = r [n] ² (2)
z [n] = s [n] ² + s [n] w [n] + w [n] ² (3)

Assuming that the noise is a zero-average Gaussian distribution, if the above output is averaged after measuring the averaging number N _av , the decision statistic can be approximated by a chi-square distribution. Depending on the presence or absence of a signal, the distribution is either central chi-square or non-central chi-square. For this reason, discrete binary hypothesis testing can be formulated for signal detection.

Here, we consider performing energy detection in the frequency domain. As described above, there is a valid reason why the energy detection is performed in the frequency domain shown in FIG. This is because the CR sensing request may include specifying the location of the center frequency of the detection signal. Therefore, in this simulation, the received signal was converted to the frequency domain using FFT and further squared. The spectrum was smoothed by averaging the output samples over the number of averaging times _Nav . Furthermore, assuming that the received signal statistics are the same, the hypothesis can be reused in the frequency domain. Finally, signal detection was performed as shown in the following equation (4).

In the above equation (4), b _z is a decision variable, and is specifically expressed as the above equation (4-1). The left side of the conditional clause in equation (4) is expressed as in equation (4-2) above. In Equation (4-2), the number of sample inputs N _sb is the number of bins for each subband regarding frequency, and z _ij is the decision statistic of the i-th frequency bin in the j-th subband. The test statistic (the test statistic after averaging) can be expressed using noise variance (σ _w ² ) and signal variance (σ _s ² ).

In this simulation, a known Neiman Pearson (NP) detector was used to obtain a hard decision. The NP detector compares the sample average z with the threshold value γ in equation (4). In using the NP detector, the false alarm level _Pfa or the predetermined detection probability _Pd was controlled by adjusting the value of the threshold γ. Therefore, to evaluate the performance of the detector with the false alarm level P _fa and detection probability P _d. Here, the false alarm level P _fa and the detection probability P _d are expressed by the following equations (5) and (6), respectively, using the above-described statistics.

Further, when both the above formulas (5) and (6) are combined, the averaging count N _av can be expressed by a false alarm level P _fa and a detection probability P _d . Here, the false alarm level P _fa and the detection probability P _d are both functions of SNR. Therefore, the averaging number N _av can be expressed as a function of SNR. Specifically, the number of averaging times N _av is scaled by φ (1 / SNR) ² .

According to the above equation, the performance of the energy detector is inferior on the side where the SNR value is small. On the other hand, if the performance is improved by increasing the number of times of averaging N _av , the sensing time will also increase. Therefore, increasing the number of times of averaging N _av does not have a desirable effect for CR.

  Next, Reference Example 2 will be described.

  In Reference Example 2, a simulation was performed using a periodic stationarity detector. Periodic stationarity is a characteristic of a signal when the average and variance of the signal show periodicity over time. A resource of the nature of periodic stationarity is the spectral redundancy built into the communication signal during modulation. Spectral redundancy is required to be minimized from the viewpoint of spectrum efficiency, but for various reasons, it is intentionally introduced in an appropriate manner during encoding and modulation. Sometimes.

  By the way, the periodic stationarity detector is a detector for testing the periodic stationarity of a signal. Such a detector can be implemented by constructing a correlation between the signal spectrum and a shifted signal spectrum. When the signal has periodic steadiness, it is known that the magnitude of the spectrum correlation function shows a peak at a specific frequency called a repetition frequency.

  Therefore, changing the modulation scheme will change the characteristics. Therefore, in this simulation, a periodic stationarity detector was used as the main detector. FIG. 4 shows the simulation results.

  FIG. 4 is a graph showing a spectrum correlation between an RF signal including noise and a binary phase shift keying (BPSK) signal (primary signal). The symbol rate of the primary signal was 5 Mb / s, and the SNR was −24 dB. In FIG. 4, the peaks on both sides of 0 MHz indicate the repetition frequency (in this case, the symbol rate and an integer multiple thereof).

  As can be seen from FIG. 4, according to this simulation, the peak size at each repetition frequency is clearly different depending on the repetition correlation coefficient. The repetitive correlation coefficient varies depending on the baseband pulse shape.

  That is, it was found that the periodic stationarity detector according to the present reference example 2 is inferior in performance as compared with the energy detection by the conventional detector described above. This is thought to be because the energy contained in the characteristics is related to the pulse shaping roll-off filter used to generate the primary signal.

  Moreover, in this reference example 2, when using a detector, it turned out that the different characteristic of a primary signal needs to be stored in the receiver side beforehand. This is because it is necessary to increase the practicality of the detector by CR in terms of computational complexity and resource usage. Note that the periodic stationarity detector can be used simply as a signal detection scheme by performing a periodic stationarity test at a specific repetition frequency. However, in this case, it is necessary to evaluate the spectrum correlation of the received signal at those repetition frequencies. In other words, in order to achieve this, there is a problem that prior knowledge needs to be prepared in advance for the repetition frequencies of all primary signals.

  Next, Example 1 will be described. In Example 1, a simulation was performed using the detector 10 as shown in FIG.

The decision statistic for binary hypotheses in detection using autocorrelation is derived from the autocorrelation sequence of the signal instead of the received signal itself. Therefore, in this simulation, first, the autocorrelation of the received signal is calculated with respect to the delay time τ (= 1,..., Τ _max ), and the obtained signal after the autocorrelation processing corresponds to the averaging number N _av . Averaged over the observation period. Here, in this simulation, the maximum value | τ _max | of the delay time τ is selected so as to be equal to or less than the sampling time T _s (that is, | τ _max | ≦ T _s ). Finally, a decision statistic was obtained by converting into the frequency domain using the FFT device 16. When the average autocorrelation of the equation (1) is expressed as a function of the delay time τ, the following equation (7) is obtained.

Here, it is assumed that the input noise processing is white Gaussian and is not correlated with the primary signal s (t). Under this assumption, when a very large averaging is performed, it can be seen that the left side of equation (7) is dominated by two terms at the rear of the right side of equation (7). In other words, when the number of times of averaging N _av approaches infinity, this corresponds to the sum of two terms in the front part of the right side of equation (7) approaching 0 (the following equation (7-1)).

In this simulation, we first considered the autocorrelation of the primary signal when using a specific modulation scheme. (Note that the modulation scheme specified here is merely an example. Any modulation scheme can be considered as long as the primary signal is narrower than the system bandwidth.) In this simulation, it is assumed that s (t) is a BPSK modulation signal that can be expressed as the following equation (8).
s (t) = (2P) ^1/2 a (t) cos (2πf _c t + θ) (8)

In Equation (8), P, f _c , and θ represent an exponent, a carrier frequency, and shift modulation, respectively. The function a (t) represents a baseband signal and is given by the following equation (9). In the formula (9), a _n is a binary stream of data, T _s is the symbol period (sampling time), q (t) is a pulse shape of the base band. In this case, the autocorrelation function is given by the following equation (10). In equation (10), the value of | τ | is equal to or shorter than the symbol period T _s (that is, | τ | ≦ T _s ) and other than zero.

Next, in this simulation, we considered the autocorrelation of noise. Specifically, the following formula (11) was obtained by taking into account the CR wideband front end and the static hypothesis and averaging. Wherein (11), N _0/2 is the input noise of the PSD (power spectral density), the [delta] _tau, which is the Dirac delta function.

Then, detection using autocorrelation in the frequency domain (hereinafter also referred to as “autocorrelation detection”) is performed in the same manner as the energy detection method using the conventional detector of Reference Example 1 (hereinafter also simply referred to as “energy detection”). The decision statistic was taken into account. In the frequency domain, a null hypothesis is obtained from the FFT output of Equation (11). When the value of | τ | is equal to or shorter than the symbol period T _s in Equation (7) (| τ | ≦ T _s ), A signal hypothesis is obtained.

Note that if the condition | τ | ≦ T _s is taken, the signal mixed with the auto-correlation of noise and the complete correlation signal in equation (10) will be processed, and hence the FFT device No. 16 partially performs asynchronous detection equivalent to sine wave detection. It is already known that such sine wave sensing has a better scaling law in energy detection than a relatively wideband signal such as QPSK.

Also, since the primary signal does not have a fixed bandwidth, the optimal selection of the delay time τ depends on the bandwidth of the primary signal concerned. If more than two primary signals are involved, T _s assumes the symbol period of the primary signal with the maximum bandwidth.

  In this simulation, the averaging process is performed after the autocorrelation process in the autocorrelation detection procedure (see FIG. 2). This has shown that complexity can be minimized.

  Further, in this simulation, the subband approach was applied in the same manner as described in Reference Example 1. Specifically, instead of deriving accurate statistics, the FFT bin size and the number of averaging used in the conventional energy detector were used as they were instead. As a result, it was found that better detection performance can be achieved. In other words, when the same input parameters are used, the autocorrelation detection according to the first embodiment improves the performance even if the complexity is partially increased compared to the energy detection by the conventional detector according to the first reference example. I understood that. Although details will be described later, the autocorrelation detector (detector 10 of the present invention) can achieve better detection performance even if the complexity level is kept at the same level. This improvement in performance can be attributed to the fact that detection problems have been effectively reduced by sine wave detection.

  In autocorrelation detection, for example, if the noise power is abandoned due to the cross term included in the signal for the noise correlation for calculating the null hypothesis parameter, the actual noise variance used to calculate the threshold is calculated. Decrease. In particular, when the primary signal is strong, the noise floor of the other subbands increases due to the power of the cross term of the additional noise, which increases false alarms in the other subbands. That is, if a signal is present and that signal is detected in one of the subbands, it will cause a false detection in the other subband. When a plurality of primary signals exist at a time, the autocorrelation detector 10 detects not only individual primary signals but also a cross term between two different primary signals. In this case, false detections may increase again in the subband where the cross term exists. Considering these points, the first embodiment has a particularly remarkable effect when the power of the primary signal is large.

  By the way, the computational complexity at the time of detection is important because it not only directly affects the entire sensing time but also directly affects the power consumption of the device. Therefore, the computational complexity was examined in more detail.

Table 1 summarizes the computational complexity of each method described above.

Here, the complexity of the calculations, the input variables of the delayed samples is accompanied using the equation L τ ₌ τ _max / t _s correlation. The input variables include the sampling time t _s of the received signal, the averaging number N _av , the number of repetition frequencies N _α , the number of sample inputs N _s , and the number of FFT points N _pt . The equations in Table 1 are obtained by adding the values of algebraic calculations involved in each type of detection method.

For simplicity, Table 2 shows an example of the computational complexity when setting the same parameters. Specifically, the correlation L _{τ is set} to 100, the averaging count N _{av is set} to 1000, the repetition frequency number N _{α is set} to 1, the sample input number N _{s is set} to 1000, and the FFT point number N _{pt is set} to 1024. The results shown in Table 2 were obtained.

In this description, the result of the periodic steadiness detector (Reference Example 2) is excluded from the comparison target. This is because the reference example 2 does not need to be compared for the reason described above. That is, Table 2 is used to compare the result of energy detection (Reference Example 1) with the result of autocorrelation detection (Example 1). And when they were compared, it was found that autocorrelation detection is more complex because it requires more operations when the same input parameters are used. On the other hand, for example, changing the number of sample inputs N _s and the number of averaging times N _av , which are input parameters, is a tradeoff with the degree of performance improvement, but it has been found that the complexity can be reduced. .

  The simulation result will be described in more detail. In order to obtain a simulation result, processing was performed as follows.

  As a sample primary signal, a BPSK signal having a symbol rate of 1 Mb / s and a center frequency of 10 MHz was generated. The generated primary signal was added to Gaussian white noise. The sampling frequency was 100 MHz. As a result, we thought that sampling of the primary signal and further emulation of the broadband wireless front end would be efficient. The system bandwidth was 50 MHz.

In order to estimate the detection probability, the maximum likelihood estimated value (MLE) was used. Each simulation was repeated 1000 times so that the estimation error was about 3% of the total when the detection probability P _d was 0.1. The estimation error is given by the following equation (12) using the number of simulations _Nsim .

  Subsequently, the entire bandwidth was subdivided into 8 subbands. As a result, each subband occupies 6.25 MHz. The length of the FFT was 1024. As a result, 64 frequency bins cover one subband. In this simulation, only one primary service is assumed to be in the second subband, and detection was performed at a specific frequency. This setup is now in a state suitable for detecting spectrum holes or idle frequency bands in the frequency band. Note that CR is used by detecting a separate signal in each subband.

FIG. 5 is a graph showing the detection probability when the false alarm probability is fixed at 0.05. The number of sample inputs N _s and the number of averaging times N _av were 1000 and 1000, respectively. In FIG. 5, a star (asterisk) marker indicates a performance plot of autocorrelation detection (Example 1). On the other hand, what is indicated by a round marker is a plot of the performance of energy detection (Comparative Example 1) by a conventional detector.

As can be clearly seen from FIG. 5, in autocorrelation detection, a significant improvement in performance is recognized. Specifically, when the detection probability P _d is 0.9, the performance is improved by about 3 dB in the region where the SNR is lower than −18 dB. Further, when the detection probability P _d is 0.1, the performance is improved by about 4 dB in the region where the SNR is lower than −18 dB.

  Moreover, when the ratio regarding the complexity of the autocorrelation detection (Example 1) to the energy detection (Reference Example 1) was calculated using the relationship shown in Table 1, it was about 19.25.

  Also, in FIG. 5, the detection probabilities obtained corresponding to subbands that are nearby and do not have a primary signal are indicated by x-shaped markers. These plots are obtained by detecting the cross term of the neglected noise for the purpose of observing the influence of the cross term for calculating the threshold value. However, due to sufficient averaging, no false positives were found. It was considered that the effect of these cross terms is visualized when the SNR is increased.

FIG. 6 is a graph showing another example of the simulation result (performance plot) obtained similarly to the simulation result shown in FIG. In this example, energy detection is evaluated by varying the values of the sample input number N _s and the averaging number N _av .

In FIG. 6, the solid line with a round marker indicates the energy detection performance when the number of sample inputs N _s and the number of averaging times N _av are 1000 and 500, respectively. It was found that when the size of the averaging process was reduced, the energy detection performance was inferior to that shown in FIG.

In FIG. 6, the plots indicated by broken lines with round markers indicate the energy detection performance when the number of sample inputs N _s and the number of averaging times N _av are 10000 and 1000, respectively. In this plot, it was found that when the detection probability P _d is 0.1, the performance is greatly improved such that the SNR is improved by 2.2 dB.

In FIG. 6, the curve indicated by the star (asterisk-like) marker shows the autocorrelation detection performance when the number of sample inputs N _s and the number of averaging times N _av are 1000 and 500, respectively. ing. This curve shows that the performance is improved by 1.5 dB compared to the energy detection performance indicated by the broken line.

  Furthermore, the complexity was calculated using the parameters used in the latter two performance curves, using the complexity formulas shown in Table 1. In this case, the complexity ratio of autocorrelation detection (Example 1) to energy detection (Reference Example 1) was found to be reduced from about 19.25 to 0.72. This indicates that the complexity can be reduced by using autocorrelation detection (Example 1) in order to improve the detection performance.

  In the above description, a plurality of signal detection methods indispensable for spectrum hole detection in a cognitive radio communication system have been discussed. That is, energy detection (Reference Example 1), autocorrelation detection (Example 1), and detection using periodic stationarity (Reference Example 2) have been described.

  In addition, the discussion on the subband detection approach showed that it is effective to directly determine the positions of occupied and unoccupied frequency bands.

  Signal detector, partial energy detection, and autocorrelation detection were evaluated. It was found that autocorrelation detection performed better than energy detection when the same parameters were set, even with increased complexity. We found that autocorrelation detection still provides superior performance compared to energy detection, even at comparable levels of complexity.

    The present invention can be suitably used in fields such as wireless communication. In particular, it can be used more suitably in the field of cognitive radio communication.

FIG. 1 is a block diagram schematically showing the configuration of the detector 10 of the present invention. FIG. 2 is a flowchart showing an operation procedure of the detector 10 shown in FIG. FIG. 3 is a diagram for explaining grouping of subbands of the FFT output. FIG. 4 is a graph showing a spectrum correlation between an RF signal including noise and a BPSK signal. FIG. 5 is a graph showing the detection probability when the false alarm probability is fixed at 0.05. FIG. 6 is a graph showing another example of the simulation result (performance plot) obtained similarly to the simulation result shown in FIG. FIG. 7 is a block diagram showing a configuration of a conventional detector.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Detector 11 Analog-digital (A / D) converter 12 Averaging process part 13 Buffer 14 Autocorrelation device 16 Fast Fourier transform (FFT) device 18 Determination device

Claims (5)

A wireless signal detection method for detecting an analog wireless signal for wireless communication with a detector (10),
The detector (10)
A buffer (12) for storing a digital signal corresponding to an energy value of an input signal input from the outside;
An autocorrelation device (14) for performing autocorrelation processing on the signal;
An FFT device (16) for performing a fast Fourier transform on the signal;
A comparator that compares a predetermined upper limit value for the energy value of noise in the background of the analog radio signal with the energy value of the signal;
A determination device (18) having the comparator and for determining whether or not an analog radio signal to be detected has been detected;
Have
Said method is:
An accumulation step (S14) for accumulating a predetermined number of samples of the digital signal in the buffer (12);
An autocorrelation step (S16) for reading out the digital signal of the predetermined number of samples stored in the buffer (12) by the autocorrelation device (14) and performing autocorrelation processing;
An averaging step (S18) for obtaining a periodic signal from the signals after autocorrelation processing obtained in each of the autocorrelation steps (S16) by repeating the autocorrelation step (S16),
FFT processing step (S20) for obtaining a spectrum by performing fast Fourier transform processing on the periodic signal obtained in the averaging step (S18 ) by the FFT device (16);
A comparison step (S22) for comparing the spectrum obtained by performing fast Fourier transform processing in the FFT processing step (S20) with the upper limit value by the comparator;
A determining step (S24) for determining whether or not the analog radio signal to be detected can be detected from the input signal by the determining device (18) using the comparison result between the spectrum and the upper limit value by the comparator. When,
Have
Thereby, information about the analog radio signal to be detected is acquired from the input signal.
Wireless signal detection method. The autocorrelation device (14) is a composite device having a plurality of shift registers, a plurality of multipliers, and a plurality of adders,
In the averaging step (S18), the composite device performs the autocorrelation process on a delay time, and averages the obtained signal after the autocorrelation process in an observation period, thereby obtaining the periodic signal. ,
The radio signal detection method according to claim 1. In the comparison step (S22), the comparator divides the frequency bins of the output of the FFT device (16) into a group consisting of a plurality of subbands, and calculates an average value of each subband in the group as the upper limit value. Compare with
The radio signal detection method according to claim 1. The program for making a computer perform the step of the radio | wireless detection method in any one of Claims 1-3 . A sensor for detecting an analog wireless signal for wireless communication,
An input unit for inputting an analog wireless signal in the wireless communication area as an input signal;
A buffer (12) for storing a digital signal corresponding to the energy value of the input signal input from the input unit;
An autocorrelation device (14) for performing autocorrelation processing on the signal;
An FFT device (16) for performing a fast Fourier transform on the signal;
A comparator that compares a predetermined upper limit value for the energy value of noise in the background of the analog radio signal with the energy value of the signal;
A determination device (18) having the comparator and determining an analog radio signal to be detected;
Have
The digital signal is accumulated in the buffer (12) by a predetermined number of samples,
The autocorrelation device (14) acquires each of the autocorrelation processes by repeatedly reading the digital signal of the predetermined number of samples stored in the buffer (12) and applying the autocorrelation process. A periodic signal is obtained from the signal after the autocorrelation processing,
The FFT device (16) obtains a spectrum by performing a fast Fourier transform process on the periodic signal obtained by the autocorrelation device (14) ,
The comparator compares the spectrum obtained by the FFT device (16) performing fast Fourier transform processing with the upper limit; and
The determination device (18) determines whether the analog radio signal to be detected can be detected from the input signal using the comparison result between the spectrum and the upper limit value by the comparator,
Obtaining information about the analog radio signal to be detected from the input signal;
sensor.

Images