JP2009267711A - Signal detector and radio communication device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a signal detector and a radio communication device having a function for identifying PC noise with a simple configuration. <P>SOLUTION: The signal detector 1 calculates a temporal correlation of phase and amplitude fluctuation of adjacent harmonics, and compares the correlation to a threshold. Further, when no complete correlation is obtained depending on an order difference of the adjacent harmonics, the signal detector confirms that a phase determined by multiplying a phase difference of the difference by the order of the higher harmonics correlates with the phase of the higher harmonics. Accordingly, the signal detector confirms that periodical peaks are the higher harmonics of the same clock. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a signal detection device and a wireless communication device for identifying electromagnetic noise.

  Wireless communication has become widely used for consumer applications, and moving images and high-definition images are generally transmitted and received using wireless communication, which dramatically increases the traffic required for wireless communication. ing. On the other hand, in order to use a finite frequency resource in wireless communication, the resource depletion problem has become serious. Several methods for solving this have been proposed, and one of them is a system called cognitive radio.

  There are generally two types of methods generally called cognitive radios, one having a plurality of existing wireless communication interfaces, a multi-mode type for distributing traffic to them, and the other having one wireless communication interface, This is a free frequency detection type that detects the surrounding radio wave usage status and uses free frequencies. Hereinafter, the cognitive radio refers to a vacant frequency detection type cognitive radio, and the cognitive terminal refers to a terminal that performs communication using the cognitive radio. The idle frequency detection type cognitive radio is a future technology, and at present, many problems to be solved remain.

  One of them is a problem of electromagnetic noise emitted by a PC or the like (hereinafter referred to as PC noise). It is assumed that the cognitive terminal is used by being inserted into a slot of the PC or built in the PC. In cognitive radio, a free frequency is identified based on the carrier sense result of the terminal itself. The PC has a large number of digital processing units such as a CPU and a bus, and their digital signals, particularly clocks, emit very large electromagnetic noise. Also, these digital signals emit not only clock frequency components and lower frequency components, but also harmonics generated by sharp rising / falling waveforms. In some cases, harmonics are observed up to several GHz with respect to a clock frequency of about several hundred MHz (see, for example, Non-Patent Document 1), and the order of the emitted harmonics reaches several tens of orders. When the cognitive terminal has a function capable of detecting the surrounding radio wave usage state with very high sensitivity, there is a possibility that these PC noises are detected and the frequency is recognized as being used.

  The PC noise described above is electromagnetic noise whose information is meaningless, and naturally there is no receiver who receives it. Therefore, it cannot be said that the frequency is utilized even if PC noise is flying. Further, the PC noise is not legally permitted to occupy the frequency, and has no priority over the frequency. The cognitive terminal does not use the frequency at which radio waves are detected in the surroundings, but it is originally intended to protect the rights of the (primary) system that has usage priority for that frequency. Therefore, even if a cognitive terminal observes radio waves at that frequency, it is not necessary to stop using that frequency if it is PC noise. However, most carrier sense functions simply detect the presence or absence of power at that frequency. Also, when the cognitive terminal is a consumer wireless terminal, it is difficult to increase the amount of processing for detecting available frequencies. For example, it is difficult to know the specifications of the primary system of all frequencies that can be used, and to analyze whether the radio wave is received and analyzed to the protocol and whether the radio wave is primary.

  On the other hand, a large number of PC noises, particularly up to about 5 GHz, are observed with microwaves that are easy to use for radio. If the frequency is avoided, the frequency that can be used by the cognitive terminal decreases.

In a wireless terminal, it is known that the clock of the digital unit of the terminal itself affects the reception of the wireless signal, and the clock frequency is controlled in accordance with the wireless frequency to be used so that the harmonics of the clock are not applied to the wireless frequency. A technique is disclosed (for example, see Patent Document 1).
JP 2006-253780 A IEICE 2007 Society Conference B-17-28

However, unlike the wireless terminal disclosed in Patent Document 1, the cognitive terminal is used by being inserted into a PC slot or built in as an attached function to the PC. Thus, it is practically difficult for a cognitive terminal used as a wireless module to control the clock of the PC body.

  Accordingly, the present invention has been made to solve such a problem, and provides a signal detection device and a wireless communication device having a function of identifying PC noise with a simple configuration.

   In order to achieve the above object, a signal detection apparatus according to the present invention includes a filter means for extracting first and second peaks from periodic peaks on a frequency included in an input signal, and a waveform of the first peak. And a correlation detection means for performing two or more types of correlation calculations on the waveform of the second peak, and comparing the correlation calculation result with a predetermined threshold value, respectively, The ratio of peak waveform fluctuation is approximately m to m + a (m is a value obtained by dividing the radio frequency of the first peak by the period indicated by the periodic peak on the frequency. M + a is the radio of the second peak. And a correlation determination unit that determines that the first peak and the second peak have the same source when the frequency can be regarded as a value obtained by dividing the frequency by the period.

   In addition, the signal detection apparatus of the present invention includes a filter unit that extracts at least three from periodic peaks on a frequency included in an input signal, and a set of two peaks having different combinations among the extracted peaks. Correlation for determining that the sources are the same for the peaks included in the set exceeding the threshold by performing one or more types of correlation calculations and comparing the correlation calculation results with a predetermined threshold value, respectively. And determining means.

In addition, the wireless communication device of the present invention receives the output of the signal detection device described above and the signal detection device, and determines that the plurality of peaks determined to have the same source are electromagnetic noise. And a frequency selecting unit that recognizes a frequency determined to be electromagnetic noise by the electromagnetic noise determining unit as an available frequency.

  ADVANTAGE OF THE INVENTION According to this invention, the signal detection apparatus and radio | wireless communication apparatus which have a PC noise detection function with a simple structure can be provided.

(First embodiment)
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. In the drawings, only parts that are essentially related to the present invention are shown, and amplifiers, power supplies, etc. that are actually necessary but are not related to the operation of the invention are omitted.

  FIG. 1 shows a typical embodiment of the present invention. The signal detection device 1 according to the present embodiment is different from each other in a periodicity detection unit 3 that detects periodicity from an input signal and a filter 4 that extracts two or more peaks from the input signal in accordance with instructions from the periodicity detection unit 3. The correlation calculation unit (1) 7 and the correlation calculation unit (2) 8 that calculate the correlation values of two peaks by correlation calculation, and the correlation values calculated by the correlation calculation units 7 and 8 are determined as threshold values. And a determination unit 9 that determines whether or not two peaks have a correlation.

The periodicity detection unit 3 detects whether there is a peak indicating periodicity in the spectrum of the input signal, and notifies the filter 4 of the frequency of the peak estimated to indicate periodicity. The filter 4 filters and outputs signals corresponding to two peaks, preferably two adjacent peaks, from the input signal. These two peaks are input to the correlation calculation unit (1) 7 and the correlation calculation unit (2) 8, respectively.

In each correlation calculation unit, different correlation values are calculated for two peak fluctuations, and the result is input to the determination unit 9. A specific correlation value calculation method will be described later.

The determination unit 9 determines whether or not the calculated correlation values exceed a predetermined threshold value, and if both of the two correlation values exceed the threshold value, the two peaks to be calculated are sufficient. It has correlation and it determines with having generate | occur | produced from the same generation source.

Specifically, the ratio of the waveform fluctuation of the first peak to the waveform fluctuation of the second peak is approximately m to m + a (m is a periodic peak on the frequency indicating the radio frequency of the first peak. The value divided by the period, m + a is a value obtained by dividing the radio frequency of the second peak by the period, and m and a are integers), and the first peak and the second peak are generated sources. Are determined to be the same. The result is passed to the next stage.

(Operation principle of the signal detection device 1)
Next, the principle by which the effect of the present embodiment can be obtained by such a configuration will be described with reference to FIG. It is an object of the present invention to identify electromagnetic noise caused by a PC clock or the like. FIG. 2 schematically shows the state of the spectrum when the PC clock radiated from the PC and its harmonics reach the vicinity of the cognitive terminal.

The PC clock has strong power at Δf which is its clock frequency. In general, the PC clock has a waveform closer to a rectangular wave than a sine wave, and many harmonics are generated. Roughly, the power is large at a frequency close to Δf, and the power decreases as the order increases. Since the clock waveform is roughly vertical, the even harmonics are small and the odd harmonics are large, but the harmonics of the order of several tens are not evenly related, and each has a small power. . A band indicated as a cognitive terminal use band in FIG. 2 is a band used by the cognitive terminal, that is, a band where the cognitive terminal performs carrier sense. Among them, there are harmonics of the PC clock.

In FIG. 2, there are eight harmonics, mth to m + 7th harmonics. The spectrum of the PC clock is usually not a complete line spectrum. The CPU clock may be intentionally subjected to frequency modulation (FM) in order to reduce electromagnetic noise. Also, even a board bus clock that does not do much like that, there is jitter and minute amplitude fluctuations, and when looking at the spectrum in detail, the spectrum has a certain width (line width) .

Note that the signal detection method performed in this embodiment can be applied to a clock multiplied by FM, but in the following, for simplicity of explanation, the clock frequency is constant without FM. Suppose that

In FIG. 2, the fundamental wave component C1 has, for example, a full width at half maximum and a line width of ΔL. This component can be expressed as:

The subscript indicates the order of the harmonics, and “1” is the fundamental wave. Note that ω = 2πΔf. The mth harmonic is

The line width is approximately mΔL. In Equations (1) and (2), χ is the magnitude of the harmonic when radiated from the PC, and is a value that varies depending on the order of the harmonic. . Therefore, the spectrum does not spread by χ.

Further, b (t) indicates the fading fluctuation of the amplitude and phase received by the propagation of the PC clock and harmonics to the cognitive terminal. b (t) is a complex number and also affects the phase. Although it is subject to temporal fluctuations depending on its frequency rather than the order, in general, fading fluctuations are sufficiently slower than the jitter of the PC clock, so that the line width is not greatly affected. It is not a term that depends on the order itself.

A (t) is the fluctuation of the amplitude inherent in the PC clock, and the magnitude depending on the order is given as a separate coefficient as χ, and can therefore be shown as a power of the order. In the equations (1) and (2), since the components constituting the line width are separated into amplitude and phase, A (t) is a real number and affects only the amplitude.

φ (t) represents the jitter component of the PC clock, which is a term that is simply proportional to the harmonic order. A phase offset is applied corresponding to the order of the harmonics, which is indicated by θ, has no temporal variation, and therefore does not affect the line width.

As described above, it can be classified into an independent term for each harmonic and a term in which the value of the fundamental wave is raised to the harmonic order. An independent term for each harmonic, such as χ or b (t), cannot be known unless the PC side knows what kind of radio wave is emitted. On the other hand, terms such as A (t) and exp (jφ (t)) that are raised to the power corresponding to the order can be known only at the receiving end by comparing harmonics. That is, it is possible to know whether the harmonics of the same PC clock are different in order.

The same notation represents the m + 1st order harmonic as follows.

(Correlation calculation 1)
In the present embodiment, harmonics of the m and m + a order (preferably a = 1) are extracted, and two or more types of correlation calculations are performed on the phases. For example, the correlation for C _m and C _{m + 1} is obtained by the following equation.

τ is an integration period. There is a range of τ for effectively obtaining the correlation, which will be described in detail later.

For example, when Δf is 100 MHz and the cognitive terminal use band is at a frequency of several GHz, m is a number such as several tens. If the expression (2) and (3) m is as large as several 10 becomes a very similar waveform change, the absolute value when the correlation of C _m (t) and C _{m + 1} (t) is "1" Get closer. Therefore, in the present embodiment, by taking these correlations, it is determined whether or not these are harmonics of the same PC clock.

However, the PC clock is not as accurate as the local oscillator for wireless communication and generally has a very large jitter, but the jitter is not always very large, and sometimes the jitter is small. Thus, it is not always guaranteed that all PC clocks have a jitter of a certain level or more.

Furthermore, there is a possibility that two signals having completely different sources are coincidentally sine waves with very narrow line widths. In these cases, since the two peaks both have very little phase and amplitude fluctuations and are close to a pure sine wave, even if they are not the same PC clock harmonic, the correlation approaches “1”. There is.

In order to avoid such a situation, in this embodiment, not only one type of correlation but two or more types are calculated. The second correlation calculation is such that the characteristics are confirmed by an operation different from that of the other correlation calculation based on the characteristics of the PC noise as shown in the equations (1) and (2). By doing so, it becomes possible to identify PC noise without accidentally mistaking two sine waves having a small line width as harmonics of the same PC clock.

Note that the expressions (1) to (3) include a term related to the carrier component jωt. This term shows a very fast change, but it only shows the center frequency of the carrier and does not contain any information. The presence of this term when calculating the correlation makes it difficult to calculate the correlation of the change. Therefore, the correlation calculation is performed after converting each peak into a baseband and removing a term related to the carrier component.

Correlation calculations are possible for several parameters. Various waveforms of phase / amplitude, such as whole waveform including all phase fluctuations and amplitude fluctuations, amplitude, amplitude converted to power or envelope, phase (exp shoulder shape or radian dimension), etc. It is possible for combination and deformation.

(Configuration of wireless transceiver)
A signal detection apparatus 1 shown in FIG. 1 is used in a cognitive radio transceiver. Here, the case where the signal detection apparatus 1 shown in FIG. 1 is applied to a cognitive radio transceiver will be described with reference to FIG.

FIG. 3 shows an outline of the configuration of the cognitive radio transceiver according to this embodiment. In the wireless communication device 11 that is a cognitive wireless transceiver, a signal received by the receiving antenna 12 is input to the wireless signal receiving unit 13. The wireless reception unit 13 includes a wireless signal conversion unit 14 and a demodulation unit 15. In the wireless signal receiving unit 13, first, the signal of the wireless signal frequency is converted into a direct current vicinity by the wireless signal conversion unit 14, digitized, and output.

An example of the wireless signal converter 14 is shown in FIG. A radio signal from the receiving antenna 12 is input to the RF unit 25. This is appropriately amplified, level-adjusted, and filtered by the RF unit 25 mainly composed of an analog circuit, and down-converted to the vicinity of the direct current by the down converter 26. Next, a necessary band is filtered by a low pass filter (LPF) 27. At this time, for example, the entire cognitive terminal band shown in FIG. 2 is filtered. The filter output is A / D converted by the A / D unit 28, converted into a digital signal, and converted into a frequency domain signal by the FFT unit 29. In this case, the filter 4 in the signal detection device 1 is a filter corresponding to the frequency domain. When the filter 4 corresponds to the time domain, the FFT unit 27 is not necessary. The parts from the down-converter 26 to the A / D unit 28 are processed by being divided into an in-phase (I) component and a quadrature (Q) component as necessary, and recombined into a complex signal after A / D conversion. Also good.

The output of the radio signal converter 14 is output to the carrier sense unit 18 and the demodulator 15 as shown in FIG. The demodulator 15 demodulates the data contained in the signal and outputs it to the upper layer. The subsequent steps are omitted because they are not related to the present embodiment. The signal output from the radio signal conversion unit 14 to the carrier sense unit 18 is branched into two and input to the usage status identification unit 19 and the signal detection device 1.

The output of the signal detection device 1 is input to the electromagnetic noise determination unit 20. The electromagnetic noise determination unit 20 receives information about the frequencies of the plurality of peaks and the presence / absence of the correlation from the signal detection device 1, and determines whether there is an electromagnetic noise based on the presence / absence of the correlation and other information. judge.

On the other hand, the usage status identifying unit 19 detects the usage status within the cognitive terminal usage band from the radio signal passed from the radio signal conversion unit 14. For example, the band is divided into grids, and power detection is performed to detect whether or not power exceeding noise is present in each grid and notify the frequency selector 21.

The frequency selection unit 21 receives the outputs of the usage status identification unit 19 and the electromagnetic noise determination unit 20, and determines that the frequency determined not to be used and the frequency determined to be electromagnetic noise can be used. Then, the use frequency of the wireless communication device 11 is selected and determined and notified to the wireless signal generation unit 23.

Data to be transmitted is input to the radio signal generation unit 23, the data input to the use frequency selected by the frequency selection unit 21 is modulated and arranged, converted into a radio signal, and radiated from the transmission antenna 22.

By adopting such a configuration, even if the wireless communication device 11 is close to the PC, high-speed communication using a wide bandwidth is possible without being limited by the frequency of the PC clock.

(Details of signal detector 1)
Hereinafter, details of each part of the signal detection device 1 will be described.

(Periodicity detection unit 3)
The periodicity detection unit 3 detects a peak frequency that appears to be periodically generated within the cognitive terminal usage band. The periodicity detection unit 3 selects two adjacent peaks from the detected peaks and estimates whether they are higher-order harmonics. When it is estimated that the selected two peaks are higher-order harmonics, the frequency detector 3 instructs the filter 4 to extract the two selected peaks. On the other hand, if it is estimated that the harmonics are not the first-order harmonics, it is possible to determine that they are not harmonics of the PC clock without performing correlation calculation.

If the peak detected by the periodicity detection unit 3 is a harmonic of the PC clock as shown in FIG. 2 and the two selected harmonics are different in the first order, both center frequencies have a frequency difference between the two peaks. Should be divisible (excluding measurement errors). The periodicity detection unit 3 estimates whether or not the two selected peaks are higher-order harmonics by verifying whether the two selected center frequencies have such a relationship.

In addition, since the power of the harmonic between the two peaks selected by the periodicity detection unit 3 is small, the two peaks may actually be second-order differences and third-order differences. In this case, the center frequency of each peak may not be divisible by the frequency difference between the two peaks. Therefore, if it is not divisible, the frequency difference is halved and ３ times and divided again. If still not divisible, the periodicity detection unit 3 assumes that the two peaks are not harmonics of the same PC clock, and does not instruct the filter 4 to extract peaks.

The reason why it is not divisible is that the two sources are not the same source, but even if they are the same source, the source is not the baseband but a subcarrier generated around a certain center frequency. There is a case. The latter is, for example, a pulse radar signal. The interval between the harmonics of the PC clock is equal to the clock frequency, which is a relatively large frequency difference, whereas in the case of a pulse radar signal, the frequency difference between adjacent subcarriers is small.

In the present embodiment, if it is not divisible at the beginning, the frequency difference is halved or １ ／ times, but if generalized, it may be 1 / integer. However, up to 1/2 times and 1/3 times are sufficient here. In the case of a frequency difference that is too small, this is more likely to be a subcarrier group generated around some center frequency than the possibility that it is a harmonic of the hundreds or thousands of PC clocks. Because it is expensive.

Even if it is determined that there is a correlation between the two peaks as a result of the correlation detection performed by the signal detection device 1 for the same reason, the electromagnetic noise determination unit 20 determines that the frequency difference between the two peaks is smaller than a predetermined value. It is determined that it is not a harmonic of the PC clock. Since the PC clock is about 80 MHz at the latest, it is preferable to set the degree as a threshold.

The periodicity detection unit 3 detects periodicity by, for example, cyclostationary (periodic stationarity) calculation. The cyclostationary calculation is performed by decomposing the received signal into frequencies and correlating the spectrum with a slightly shifted spectrum in the frequency direction. Since the period on the spectrum that can be detected changes by changing the shift amount, various periods on the spectrum are detected by scanning the shift amount. Further, when calculating the cyclostationary, the periodicity is detected by detecting the correlation in an appropriate integration period while the signal Fourier-transformed on the frequency axis remains a complex number, and at the same time, one of the two types of correlation detection. The types can be completed at the same time, and the amount of calculation for correlation can be reduced.

As another method of detecting periodicity, for example, only a portion having a shape close to a line spectrum may be extracted from the input frequency domain signal. Moreover, after extracting the part which has a shape close | similar to a line spectrum, you may perform the above cyclostationary calculations in the unit of the frequency difference of several extracted line spectrum.

By performing the cyclostationary calculation with the complex number of the Fourier transform result, it is possible to perform up to one of the correlation calculations, but if the requirements for the integration period of the periodicity detection and the correlation calculation do not match, the cyclostationary calculation will detect the periodicity. If you are specialized, you can do it in the dimensions of power and amplitude.

In this way, periodicity is detected, and two adjacent peaks are selected from combinations of peaks that are considered to exhibit periodicity, and the subsequent correlation calculation processing is performed. When performing correlation determination, if the correlation calculation is performed completely independently for each combination of two peaks, the periodicity detection unit 3 uses a signal detection device to display a set of peaks that are considered to exhibit the same periodicity. 1 is output to the next stage together with the output of the determination unit 9. When performing correlation determination, if a peak that seems to indicate the same period is used and some processing is omitted, the peak set is passed to the next stage and correlation calculation is performed. -Notify each part that performs the judgment.

Further, the periodicity detection unit 3 detects the frequency of each peak with as high accuracy as possible. Even when the signal input to the periodicity detection unit 3 is converted into the frequency domain by FFT in the previous stage, the calculation is performed so that the accuracy equal to or lower than the resolution of the FFT is obtained.

For example, if a set of multiple peaks that are considered to exhibit the same periodicity is determined, the frequency difference between the peak at the highest frequency and the peak at the lowest frequency is divided by the number of periods contained therein. The frequency of each peak is obtained by multiplying the frequency of the lowest frequency peak by the number of cycles. At this time, for the peak at the lowest frequency and the peak at the highest frequency, these peaks can be obtained at a high frequency resolution by making the FFT period extremely long or measuring the signal over a number of FFT periods. Find the frequency of.

In this way, in most cases, the frequency of each peak is not an integral multiple of the FFT bin width. The deviation from the integer multiple becomes an offset from the center frequency of the FFT bin of each peak. These offsets are corrected when correlating, and the peak is converted to baseband before integration. In most cases, a perfectly correct offset is not sought and there is some deviation from the baseband. The deviation needs to be reduced to such an extent that it does not become a problem within the integration period at the time of correlation calculation. Alternatively, it is necessary to set an integration period that does not cause a shift. Since there are other parameters regarding the setting method of the integration time, details will be described later.

Note that the periodicity detection unit 3 is not essential in the configuration of the present embodiment. For example, if a configuration for receiving a clock frequency notification from a PC directly connected to the wireless transceiver 11 including the signal detection device 1 according to the present embodiment is used, the periodicity detection unit may be omitted. In this case, a control unit for calculating the harmonic frequency and notifying each unit of the peak frequency is required instead.

Further, the periodicity detection unit 3 does not need to detect periodicity every time carrier sense is performed, and a method of correcting the period and frequency by diverting the past results and sometimes performing detection again may be used. In particular, with respect to a peak determined to have a correlation by the determination unit 9 or determined to be electromagnetic noise by the electromagnetic noise determination unit 20, the procedure is omitted by using the past result and the primary is overlapped. It is desirable to verify whether or not

(Filter 4)
The filter 4 extracts two peaks instructed from the periodicity detection unit 3. When the input signal is a time signal, the two peaks are filtered using a narrowband filter such as FIR. If the input signal is an FFT signal in the frequency domain, the FFT bin including the peak is extracted. At that time, if the center frequency of the notified peak is the end of the FFT bin, a plurality of bins including the peak are extracted.

When overlap FFT is performed by overlapping a part of the period during FFT, the correspondence is slightly different. In this case, the ratio of the FFT period divided by the FFT interval (interval of successive FFT frames) (hereinafter referred to as the oversample rate for convenience) multiplied by the FFT bin interval, and the transmission of the window function waveform during the FFT The smaller of the function bandwidths is the effective bandwidth. If the entire line width (ΔmL) of one peak is included in the effective bin bandwidth, it is not necessary to fetch adjacent bins. If one peak protrudes from the effective bin bandwidth, adjacent bins or any bin separated by a frequency within the effective bin bandwidth may be extracted simultaneously. Of course, even in the case of overlap FFT, the same extraction method as that in the case of no overlap may be taken without considering the oversample rate.

When taking out the FFT bin, if there is no overlap, it is the same as extracting with a rectangular filter. If this is a problem depending on the processing method, an appropriate filter frequency characteristic shape may be defined, and this bin may be multiplied and filtered to synthesize those bins.

In the time domain, even if the peak is not dropped to a complete baseband, the input frequency to the filter (hereinafter referred to as the intermediate frequency (IF) for the sake of convenience) remains unchanged during the correlation calculation. The correlation calculation may be performed after frequency-shifting one of the frequencies. Of course, the correlation calculation may be performed after dropping the two peaks from the respective IFs to the baseband.

(Offset compensation)
Hereinafter, a method for compensating for a frequency offset after extracting a related frequency by the filter 4 in the case of the configuration by FFT will be described. Due to the nature of FFT, the center frequency of each bin is automatically converted to baseband. Therefore, first, for each peak, it is calculated how far away the IF is from the center frequency of the bin containing the peak. As described above, since the periodicity detection unit 3 calculates the center frequency of the peak with sufficient resolution, it can be easily calculated by subtracting an integer multiple of 1 FFT bin width from the result. Next, assuming that the frequency of the offset from the center frequency is fo (fo includes a positive or negative sign), the bin is multiplied by exp (-2πfo · t). t is the time at each point of the integration period (the time of each FFT frame in the case of the FFT configuration). This causes a frequency shift that is downconverted to the bin center frequency.

When a peak extends over two or more bins, the same processing as described above is performed for one bin mainly including the peak. For adjacent bins, let fb be the difference between the center frequency of the bin containing the peak and the adjacent bin (fb contains a positive or negative sign, so fb is a sign that depends on the FFT bin width and the position of the adjacent bin. Multiplied by exp [−j2π (fo−fb) t]. The multiplication result may be added to the frequency offset compensation result of the bin mainly including the peak.

In the case of performing overlap FFT, the center frequency of each FFT bin is different for each bin depending on the oversample rate. Therefore, the center frequency of each bin is not in the direct current of each bin after FFT. Since the signals having various IF frequencies are in the frequency obtained as a result of the remainder calculation of the IF frequency by the FFT bin width × oversampling ratio in each bin, the same signal exists at the same frequency in adjacent bins. become. Therefore, in most cases, frequency offset compensation is not necessary in the synthesis of adjacent bins, and only the phase offset is compensated and the frequency is added as it is. Since the phase offset is fixedly generated between adjacent bins depending on the shape of the window function used for the FFT, one of the values is corrected with the value, and the corrected signal is added to the other. The two peaks of the results obtained in this way have their center frequencies shifted by the result of the remainder calculated by dividing the period by the FFT bin interval x the oversampling rate. Correlation processing may be performed. The frequency offset is applied as described above.

  In the configuration using the overlap FFT, there is a case where the peak exists over the frequencies at both ends of the FFT result, and it is difficult to perform the correlation process. In that case, the correlation calculation may be performed after the frequency is shifted once to a frequency at which the calculation is easy to perform. In this case, of course, the frequency shifts accordingly, so that the correlation calculation is performed after compensation.

(Correlation calculation)
Next, a specific method of correlation calculation will be described. FIG. 5 is a diagram showing an example thereof. As described above, the two peaks extracted by the filter 4 and converted into the baseband are respectively input to the complex conjugate multiplier 5. In FIG. 5, the correlation calculation unit (1) is the correlation between two peaks as shown in the equation (4). The complex conjugate multiplication unit 5 calculates C _m (t) C ^* _{m + 1} (t), which is the content of the numerator integral of Expression (4). There is no integration at this stage. Here, C ^* _{m + 1} (t) means a complex conjugate of Cm _{+ 1} (t).

At that time, C may include fluctuation components of all the amplitudes and phases as in the equations (2) and (3), or may be only the phase. The correlation calculation unit (1) 7 integrates the output of the complex conjugate multiplication unit 5 over a predetermined period based on the equation (4), performs normalization corresponding to the denominator, and calculates the correlation value. In the case of a phase-only operation, the denominator calculation is meaningless, and instead it is normalized by dividing by the time corresponding to the integration period or the number of samples. As described above, when the value of m is a very large value of several tens, if the two peaks are harmonics of the same PC clock, the absolute value of the correlation calculation result approaches “1”. In this case, depending on the order, the correlation of “−1” may be approached, and the correlation calculation result is determined by an absolute value.

In the correlation calculation unit (2) 8, only the phase of the complex conjugate multiplication result is extracted, while only one phase of the two peaks output from the filter 4 is extracted. Here, for the sake of convenience, the m-th peak is used. Although a is basically “1” in the present embodiment, it may of course be “−1”, so that generality is not lost even if it is limited to the m-th order peak. As described above, a may be “2” or “3”, and of course may be “−2” or “−3”.

The phase rotation of the complex conjugate multiplication result is multiplied by m / a, and the phase of the m-th peak is correlated with a predetermined integration period. The principle will be described with reference to FIG. 6 by taking the case of a = 1 as an example. In the correlation calculation unit (1) 7, if m is sufficiently large as described above, the absolute value of the result approaches approximately “1”, but there is a difference for the first order, so that “1” is completely set to “1”. There is always a part where the correlation cannot be taken. As shown in FIG. 6, the phase of the portion where the correlation cannot be completely obtained is equal to φ (t) which is the phase component of the fundamental wave (difference between (m + 1) φ (t) and mφ (t)). Therefore, if this is multiplied by m, it should be equal to mφ (t). If a = 1 is not true, the difference is aφ (t). Therefore, if this is multiplied by m / a and correlated with the m-th order peak, the result is ideally “1”, and it does not become “1” completely due to phase rotation due to noise, distortion, fading, etc. It should approach “1”.

When the result of the correlation calculation unit (1) 7 approaches “1”, in addition to the case where the two peaks are harmonics of the same PC clock, both are independent signals and happen to be very close to a sine wave. A case where the fluctuation is small is considered. If these are perfect sine waves, the result of the correlation calculation (2) will be close to “1”. However, no matter how small the signal is, there is always jitter. The jitters of two independent signals are usually independent, and by multiplying them by m, the difference in jitter is multiplied by m. Since the two signals are independent, the probability that the difference between the difference m times and the original phase is close to “1” is very small. Therefore, the correlation calculation unit (2) 8 can confirm the result of the correlation calculation unit (1) 7 more firmly.

When the periodicity detecting unit 3 confirms that the line widths of the two peaks are small, the correlation calculation (1) is omitted because “the line widths of both are narrow”. Only the correlation calculation (2) may be performed.

In the description of the method shown in FIG. 5 above, the phase is calculated in a form on the shoulder of exp. At this time, if the operation of m times the phase is not completely m but a number close to m, the correlation calculation result is a value close to “1” as in the case of the correlation calculation unit (1) 7. Therefore, a number close to m is sufficient. In addition, it is clear that the same correlation calculation can be performed within the range of application by simple four arithmetic operations, such as multiplying the phase of one peak by k / m and multiplying the phase of the difference by (m−k) / a. .

As another method, FIG. 7 shows an example in which a correlation is made with a radian dimension extracted only from the shoulder from exp. From the output of the filter 4, the phase fluctuation extraction unit 31 extracts only the phase in the radians dimension. At that time, in a general calculation, the phase is output within a certain range (for example, ± π), and takes a discontinuous value when exceeding the end. This becomes an obstacle when calculating the correlation. Therefore, it is preferable that the phase range be extended so that such discontinuity does not occur. In this case, the correlation calculation unit (1) 7 takes the correlation between mφ (t) and (m + a) φ (t) in FIG. 6 as they are. Since these are only a fixed number m or m + a, the correlation is ideally “1” if the harmonics are the same PC clock. Of course, the threshold value is determined because it does not become “1” completely due to noise or the like but only approaches “1”.

In this case, since the correlation calculation is ideally “1”, the operation on the difference as shown in FIG. 5 is meaningless. Therefore, only the amplitude fluctuation is extracted from the output of the filter 4 by the amplitude fluctuation extracting unit 32. The correlation calculation unit (2) 8 takes correlation only for the amplitude fluctuation portion. In this case, it is not completely “1” due to the order difference a, but if m is sufficiently large and a is small, it approaches “1”. Similarly, the threshold value is determined.

As for the method of taking out from the shoulder of exp so as not to take a discontinuous value, differentiation is performed in a state of riding on the shoulder of exp of amplitude 1, that is, a complex number state consisting of a real part and an imaginary part. This is possible by taking out the amplitude component that appears.

Next, a method for extending the method of FIG. 5 to verify whether or not three or more peaks are harmonics of the same PC clock will be described. First, correlation calculation is performed on two adjacent ones in the same manner as in FIG. Next, the order of the third peak is m + b (b ≠ 0, b ≠ a). The phase of the complex conjugate multiplication result obtained in the first calculation is multiplied by (m + b) / a, and the correlation with the phase component of the third peak is calculated. As long as the third peak is not correlated with the first two peaks, the result does not approach “1”. Note that the correlation between the first two peaks needs to be confirmed before entering the verification of the third peak. If no correlation is confirmed in the first two peaks, the next step is not entered until a pair that can be confirmed is found. Furthermore, when the threshold of the correlation between the first two peaks is determined, even if it is determined that there is a correlation at the limit of the threshold, a pair showing a stronger correlation is separately searched for, and the complex conjugate multiplication result is obtained. Find it and use it in the next step. In this way, if it is sequentially applied to other peaks in the same set, the correlation calculation unit (2) 8 can be omitted.

When the correlation calculation unit (2) 8 is omitted, there is a method of verifying whether or not the phase differences of two adjacent combinations of a plurality of peaks are all the same as shown in FIG. Description of the same parts as other figures is omitted. The filter 4 is notified of a set of peaks estimated to have the same generation source from the periodicity detection unit 3, and filters them. The phase difference detection unit 33 calculates a phase difference for a set of these two adjacent peaks. Next, the correlation calculation unit (1) 7 calculates the correlation between the detected phase differences of different groups, and notifies the determination unit 9 of the result. Specifically, it is as follows.

When the complex conjugate multiplication result of the first two peaks (m-th order and m + a-th order, for example, a = 1) is obtained, the third peak (m + b-th order, for example, b = -1) and one of the first two peaks A complex conjugate multiplication result with the (m-th order) is obtained, and correlation calculation is performed for the phases of the obtained two complex conjugate multiplication results while eliminating the order difference. That is, the correlation calculation is performed after the phase of the former complex conjugate multiplication result is multiplied by 1 / a and the phase of the latter complex conjugate multiplication result is multiplied by 1 / b. If the result is close to “1”, it is determined that the three are correlated.

At that time, for example, if the m-order harmonic is a sine wave with very small jitter, the correlation is obtained, for example, in a frequency relationship such that a = 1 and b = −1. Sometimes. Therefore, another correlation calculation used together is, for example, comparison of line widths of three peaks, and the line widths are preferably almost the same. Alternatively, a correlation calculation between the m-th order and one of the remaining two may be performed.

When applying to a plurality of peaks existing at equal intervals, the complex conjugate multiplication results are obtained for all two adjacent peaks, and the complex conjugate multiplication results of the adjacent sets including each of the peaks included in each set are obtained. Calculate whether correlation is possible. If the correlation is obtained on both sides, it can be determined that there is a correlation for the four peaks including them. When correlation is obtained with only one of the peaks, it is determined that the peak that is included in the pair that could not be obtained and that is not included in the middle pair has no correlation. If it is not possible on both sides, doubt whether the middle pair is uncorrelated. It is better to find a boundary with / without correlation from other combinations of peaks with correlation, and judge from where the correlation cannot be obtained.

As another method, a set of a plurality of peaks is divided into two sets, and the complex conjugate multiplication results are obtained. Then, it is detected whether the phases of the complex conjugate multiplication results are correlated. In this case, there is a possibility that the correlation between two different signals whose periods are almost the same and happens to be half the frequency of each other may have a value close to “1”. It is better to use the correlation calculation inside.

(Integration time)
The integration time for the correlation calculation is determined according to the following criteria. First, the line width of each peak is measured in advance, and when the order of the order is set, the line width (FWHM: Full Width Half Maximum) is divided by the order, and the reciprocal is obtained to obtain the time dimension. This is the speed of fluctuation of the component that is considered to be a fundamental wave. When the time integration corresponding to this time is integrated, there is no problem in the correlation calculation in which the correlation is ideally “1”, but the first order difference as in the correlation calculation unit (1) 7 in FIG. 5 is ignored. In such a method, the correlation value is significantly lowered. Therefore, the determined time is multiplied by a certain constant of about 1/10 to 1/4. This is temporarily called the first time. The first time is valid only in a manner that ignores the primary difference. On the other hand, as shown in equations (1), (2), and (3), there are components that vary due to fading and components that vary due to Doppler shift. Wireless systems always specify the maximum amount of Doppler shift that they can handle. Fading can be seen as a different aspect of the Doppler shift caused by the movement of the terminal, and the fading speed can be roughly estimated from the maximum Doppler shift amount defined by the system. When adjacent peaks are separated by a unit of 100 MHz, the influence of fading on these can be regarded as independent. Therefore, unless the integration time is made sufficiently shorter than the time corresponding to the reciprocal of the fading speed, the correlation decreases due to the influence of fading. Therefore, the time determined from fading is multiplied by a constant constant of about 1/10 to 1/4. This is temporarily called the second time. Note that if the phase variation due to fading is compensated at any time by an equalizer or the like, the second time may be set to a sufficiently long time or determined according to the amount of no compensation partition.

Finally, as described above, there is a shift corresponding to the detection error of the center frequency of each peak. The maximum value of the detection error is a fixed numerical value depending on the frequency detection method. The maximum value of the error is doubled considering the case where both peaks are shifted by the maximum value in the opposite direction. It is reciprocally converted into time, and this is similarly reduced to about 1/10 to 1/4. This is assumed to be the third time.

Of the first to third times calculated so far, the shortest time is set as the integration period. When the carrier sense period is very short and the correlation calculation cannot be performed sufficiently (for example, when the FFT is only about a few frames and the concept of time is not meaningful), a plurality of skipped carrier sense periods You may straddle. At that time, care is taken not to read discontinuities due to the interruption of the period, and the total integration period should not exceed the predetermined integration period. In addition, the peak frequency is detected with sufficient accuracy, that is, with sufficient time to obtain sufficient resolution, so that the accuracy of the peak frequency is very low and the correlation calculation cannot be performed with a sufficient length. It is desirable to detect them.

(Determination unit 9)
Next, threshold determination is performed by the determination unit 9, and the threshold is determined corresponding to each correlation calculation method and period. Also, if possible, the peak SNR (Signal to Noise Ratio) extracted by the filter is detected by the periodicity detection unit 3, the filter 4, etc., and the threshold value is adjusted accordingly. Most methods ideally have a numerical value close to “1” or “1”, but the correlation value decreases due to noise, distortion, phase variation due to fading, frequency offset error, and the like. In particular, since the influence of SNR degradation is great, it is preferable to first determine a threshold value due to other factors and change the threshold value in accordance with the SNR. The SNR may be determined based on the thermal noise at each frequency of the receiver that has been measured in advance and the measured peak power.

When two or more correlation calculation results both exceed the respective threshold values, the determination unit 9 determines that the related peaks are correlated. When the determination unit 9 or the electromagnetic noise determination unit 20 divides the threshold at a point in time when there is a set of peaks that have exceeded the threshold in the past, particularly in a system that does not detect the peak frequency every time, the center frequency I doubt the detection error. Therefore, the peak line correction unit 3 is instructed to correct the peak frequency, and the correction is performed. If it still falls below the threshold value, it is estimated that the correlation has dropped due to factors such as the primary signal overlapping one of the peaks, and it is determined that there is no correlation.

As described above, the correlation value is determined based on an absolute value, particularly for a correlation value of a calculation method in which the a-order difference is directly correlated. Even if the primary does not overlap exactly with the peak frequency, if the power of the filter is used, or the bin width of the FFT, or the bandwidth determined by the FFT window function and the overlap FFT, It is detected in the form of a decrease in the correlation value in the manner of determination as described above. In this case, since the cognitive terminal cannot transmit at that frequency, it should be determined that there is no correlation. According to the configuration of the present application, the request can be satisfied.

  Finally, the PC clock may be subjected to FM modulation to suppress electromagnetic radiation. Even in such a case, the present application can be applied with substantially the same configuration. However, it may span more than two FFT bins. The method of synthesis is almost the same as in the two cases. If there is no oversampling, the amount of frequency offset can be changed according to the number of bins up to the synthesis destination. When there is an overlap FFT, if the bandwidth of the signal resulting from FM modulation does not exceed the FFT bin interval x the oversampling rate, an additional offset is added to the center frequency so that it will not be folded at the end of the band. It is enough to hang. When exceeding, it is necessary to combine adjacent bins or bins selected at an appropriate interval corresponding to the oversample rate in the same manner as when there is no overlap. In this case, the overlap ratio and the shape of the window function are selected so that the effective bandwidth of one bin determined by the FFT window function is smaller than the FFT bin interval x the oversample rate so that signal aliasing does not become a big problem. There is a need to. Also, if the integration time is too short due to the line width of the fundamental wave, the integration time is ideally not used in the correlation calculation method, which ignores the primary difference. It is better to use a method of “1”.

  In addition, this invention is not limited to the said Example as it is, A component can be deform | transformed and embodied in the range which does not deviate from the summary in an implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiments. Furthermore, constituent elements over different embodiments may be appropriately combined.

It is typical embodiment of the signal detection apparatus of this invention. It is a figure for demonstrating the principle of this invention. It is typical embodiment of the radio | wireless communication apparatus of this invention. It is a figure for demonstrating a part of radio | wireless communication apparatus of this invention. It is the figure which showed one of the embodiment of the signal detection apparatus of this invention. It is a figure for demonstrating the operation | movement of the correlation detection of this invention. It is the figure which showed one of the embodiment of the signal detection apparatus of this invention. It is the figure which showed one of the embodiment of the signal detection apparatus of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Signal detection apparatus 3 ... Periodicity detection part 4 ... Filter 5 ... Complex conjugate multiplication part
6 ... phase adjustment unit 7, 8 ... correlation calculation unit 9 ... determination unit 11 ... radio communication device 12 ... receiving antenna 13 ... radio signal receiving unit 14 ... radio signal conversion Unit 15 ... demodulator 18 carrier sense unit 19 usage status identification unit 20 electromagnetic noise determination unit 21 frequency selection unit 22 transmission antenna 23 radio signal generation Unit 25 ... RF unit 26 ... down converter 27 ... low-pass filter 28 ... A / D unit 29 ... FFT unit 31 ... phase fluctuation extraction unit 32 ... amplitude fluctuation extraction Unit 33 ... Phase difference detection unit

Claims (15)

Filter means for extracting first and second peaks from periodic peaks on the frequency included in the input signal;
Correlation detecting means for performing two or more predetermined correlation calculations for the first peak waveform and the second peak waveform;
By comparing each correlation calculation result with a predetermined threshold value, the ratio of the first peak waveform fluctuation to the second peak waveform fluctuation is approximately m to m + a (m is the radio frequency of the first peak. The value obtained by dividing the period by the periodic peak on the frequency, where m + a is the value obtained by dividing the radio frequency of the second peak by the period), and the first peak and the second peak. Is a correlation determination means for determining that the source is the same, and
A signal detection device comprising: Filter means for extracting at least three from periodic peaks on the frequency included in the input signal;
Among the extracted peaks, one or more types of correlation calculation is performed for each pair of two peaks having different combinations, and the result of the correlation calculation is compared with a predetermined threshold value to obtain a set that exceeds the threshold value. Correlation determining means for determining that the included sources are the same for the included peaks;
A signal detection device comprising: 3. The signal detection apparatus according to claim 1, further comprising periodic peaks on the frequency of the input signal and periodicity detecting means for detecting the period.   4. The signal detection apparatus according to claim 3, wherein the periodicity detection means detects a periodic peak and a period of the input signal by calculating a cyclostationary characteristic. The correlation determination means includes
Complex conjugate multiplication means for performing a complex conjugate multiplication of one of the first peak and the second peak after frequency offset compensation and calculating a complex conjugate multiplication result;
First correlation detection means for calculating a first correlation value by integrating the complex conjugate multiplication results;
Second correlation detection means for calculating a correlation value between the phase component of the complex conjugate multiplication result m / a times and the phase component of the first peak as a second correlation value;
The correlation determination means includes
When the first correlation value is equal to or greater than a first threshold and the second correlation value is equal to or greater than a second threshold, the first peak and the second peak are generated from the same source. The signal detection device according to claim 1, wherein:   6. The signal detection apparatus according to claim 5, wherein the multiplication of the filter characteristics and the conversion to the baseband are performed by extracting bins including the first and second peaks from the FFT result.   The apparatus further comprises means for detecting an offset between the first or second peak frequency and the center frequency of the FFT bin, and means for correcting the offset at the time of conversion to baseband. The signal detection device according to claim 6. The signal detection apparatus according to claim 1, wherein a is a = 1. 3. The signal detection apparatus according to claim 2, wherein two peaks included in one set are adjacent to each other. 4. The signal detection apparatus according to claim 3, wherein when m and a cannot be regarded as an integer including a detection error, m and a are recalculated by setting the period to 1/2 or 1/3. 11. The method according to claim 10, wherein, as a result of the recalculation, when the m and the a cannot be regarded as an integer including a detection error, the first and second peaks are determined not to have the same source. Signal detection device The filter means further extracts a third peak,
The second correlation detection means further calculates (m + b) / a times (m + b (b is an integer)) of the phase component of the complex conjugate multiplication result by the first and second peaks, A value obtained by dividing a peak radio frequency by the period) and a third correlation value by the phase component of the third peak;
The correlation determination means determines that the source of the third peak is the same as the first and second peaks when the third correlation value is equal to or greater than a third threshold. The signal detection device according to claim 5, wherein The three peaks extracted by the filter means are defined as first, second, and third peaks, respectively, and a value obtained by dividing the radio frequency of the first peak by the period indicated by the periodic peak on the frequency is m. The value obtained by dividing the radio frequency of the second peak by the period is m + a, and the value obtained by dividing the radio frequency of the third peak by the period is m + b.
The complex conjugate of one of the first peak and the second peak and the other multiplication are performed after frequency offset compensation to calculate a first complex conjugate multiplication result, and the first peak and the third peak are calculated. Complex conjugate multiplication means for performing multiplication of the complex conjugate of one of the peaks and the other of the peaks after frequency offset compensation to calculate a second complex conjugate multiplication result;
A fourth correlation value that is a correlation between a value obtained by multiplying the phase component of the first complex conjugate multiplication result by 1 / a and a value obtained by multiplying the phase component of the second complex conjugate multiplication result by 1 / b. Correlation detection means for calculating
A determination unit that determines that the first, second, and third peaks have the same generation source when the fourth correlation value is equal to or greater than a fourth threshold;
The signal detection device according to claim 2, further comprising: The filter means extracts the first to fourth peaks, m is a value obtained by dividing the radio frequency of the first peak by the period indicated by the periodic peak on the frequency, and the radio frequency of the second peak M + a, a value obtained by dividing the third peak radio frequency by the period n, a value obtained by dividing the fourth peak radio frequency by the period n + b,
The complex conjugate of one of the first peak and the second peak and the other multiplication are performed after frequency offset compensation to calculate a first complex conjugate multiplication result, and the third peak and the fourth peak are calculated. Complex conjugate multiplication means for performing multiplication of the complex conjugate of one of the peaks and the other of the peaks after frequency offset compensation to calculate a second complex conjugate multiplication result;
Correlation detection for calculating a fifth correlation value that is a correlation between 1 / a times the phase component of the first complex conjugate multiplication result and 1 / b times the phase component of the second complex conjugate multiplication result Means,
A determination means for determining that the first, second, third, and fourth peaks have the same source when the fifth correlation value is equal to or greater than a fifth threshold;
The signal detection device according to claim 2, further comprising: A signal detection device according to claim 1 to 14,
Electromagnetic noise determination means for receiving the output of the signal detection device and determining that a plurality of peaks determined to have the same source are electromagnetic noise;
A radio communication apparatus comprising frequency selection means for recognizing a frequency determined as electromagnetic noise by the electromagnetic noise determination means as an available frequency.

Images