JP5004991B2 - Intruder identification device - Google Patents

Description

  The present invention relates to an intruder identification apparatus that uses radio waves to identify that an intruder has entered a detection area.

  2. Description of the Related Art Conventionally, for example, various devices that detect an intruder that has entered a detection area such as a track have been adopted. Examples of the intruder include a vehicle, an intruder, and an animal. In terms of crime prevention, the most important intruder is an intruder (human) who may harm the equipment in the detection area. An intrusion detection apparatus described in Patent Document 1 includes a transmission unit that supplies a transmission signal to a transmission antenna, a reception antenna that receives a radio wave radiated from the transmission antenna, and a reception signal received by the reception antenna. Demodulating means for demodulating the real part component and the imaginary part component, and converting the complex signal consisting of the real part component and the imaginary part component demodulated by the demodulating means into frequency data, and forming a positive frequency constituting the converted frequency data Detecting means for determining the presence of an intruding object when the difference value between the negative frequency and the negative frequency is greater than a predetermined threshold value. This prevents false detection caused by the linear change of the complex signal caused by the fluctuation of the reflectance of the ground or the like due to the shaking of the transmitting antenna or the receiving antenna due to the wind or rain.

  Further, the intruder detection method described in Patent Document 2 is an intruder detection method for detecting an intruder in a detection area using a transmission unit that radiates radio waves, and is different from the intruder. Scanning in different directions from the transmission means or detecting the radio waves from the transmission means provided at a plurality of different positions so that the area irradiated on the intruder differs by irradiating from the direction. A transmitting step for transmitting into the area, a receiving step for receiving the reflected wave of the radio wave transmitted by the transmitting step, and a measuring step for measuring the received wave intensity based on the reflected wave received by the receiving step And based on the received wave intensity measured by the measurement step according to the difference in the size of the area irradiated from different directions in the transmission step. Te, wherein the intruder and a person determining step of determining whether a human body.

Japanese Patent No. 3878103. Japanese Patent No. 3521899.

  However, according to the intrusion detection device according to Patent Document 1 and the intruder detection method according to Patent Document 2, an alarm that informs the intruder of intrusion may be erroneously issued even when the radio wave is caused by a natural phenomenon. For example, in Patent Document 1, there is a problem that a false alarm is generated when a change in the difference value due to changes in the natural environment due to wind, rain, or the like is larger than the threshold value. Further, in Patent Document 2, when the received wave intensity is within a predetermined range, it is determined that the intruder is a human body. However, when the magnitude of fluctuation of the received wave intensity due to the intrusion of the human body varies. Therefore, there is a problem that it is necessary to set the above range relatively wide, and erroneous reports are likely to occur.

  An object of the present invention is to solve the above-described problems and to provide an intruder identification apparatus capable of accurately identifying that an intruder has intruded compared to the prior art even when the environment changes.

An intruder identification device according to the present invention generates a predetermined transmission signal and wirelessly transmits it using a transmission antenna, wirelessly receives the transmitted transmission signal using a reception antenna, and receives the wireless reception in the intruder identifying device provided with a received signal and receiving means for quadrature detection to be demodulated into complex demodulated signal using the transmission signal the occurrence, in each of the plurality of analysis periods having a predetermined period length, the complex Based on the extracted feature quantity pattern, a feature extraction means for extracting a feature quantity pattern that changes in the analysis period and changes in a different manner between the intruder and the non-intruder from the demodulated signal. And whether or not the intruder has intruded is identified using at least one of a reference feature amount pattern that indicates when the intruder has entered and a reference feature amount pattern that indicates when the intruder has not entered. Characterized in that a discrimination means for outputting an identification signal indicating the identification result.

  In the intruder identification apparatus, the feature amount pattern may include at least one of a size and a direction of a plurality of vectors representing a locus when the complex demodulated signal changes discretely on a complex plane in each analysis period. It includes a time series pattern of change.

  In the intruder identification apparatus, the feature amount pattern includes a frequency spectrum pattern of a complex demodulated signal in each analysis period.

  Further, in the intruder identification apparatus, the feature extraction means performs Fourier transform or wavelet transform on the complex demodulated signal into the frequency spectrum in each analysis period.

  Still further, in the intruder identification apparatus, the feature amount pattern includes a time series pattern of a change in covariance matrix or a change in eigenvalue of the complex demodulated signal in each analysis period.

  In the intruder identification device, the feature amount pattern may be a change in the length of a locus when the complex demodulated signal changes discretely on a complex plane in each analysis period, or a region surrounded by the locus. It includes a time-series pattern of changes in the area of.

  Further, in the intruder identifying apparatus, the transmitting antenna and the receiving antenna are each an array antenna.

  Still further, in the intruder identification apparatus, each of the array antennas is a leaky coaxial cable.

  In the intruder identification apparatus, the transmission means may perform spectrum spread on the generated transmission signal using a PN code signal and transmit the transmission signal.

  Further, in the intruder identification device, the receiving means delays the PN code signal by a plurality of different delay times to generate a plurality of delayed PN code signals, and the received signal is converted into the plurality of delayed PN code signals. A plurality of despread received signals by despreading using, and each of the plurality of despread signals is quadrature detected using the generated transmission signal, demodulated into a complex demodulated signal, and output, The intruder identification apparatus includes a plurality of the feature extraction units provided corresponding to the plurality of delay times, a plurality of the identification units provided corresponding to the plurality of delay times, and the identifications, respectively. And a position determining means for determining the position of the intruder based on an identification signal from the means.

  The intruder identification device according to the present invention changes in each analysis period from a complex demodulated signal in a plurality of analysis periods each having a predetermined period length, and is different between an intruder entry time and a non-intrusion time. A feature extracting means for extracting a changing feature amount pattern, a reference feature amount pattern representing the time of intrusion of the intruder based on the extracted feature amount pattern, and a reference feature amount pattern representing the time of intrusion of the intruder And identifying means for identifying whether or not the intruder has entered using at least one of them, and outputting an identification signal indicating the identification result. By extracting the feature amount pattern, it is possible to accurately identify that the intruder has intruded as compared with the prior art.

It is a block diagram which shows the structure of the intruder identification apparatus 1 which concerns on Embodiment 1 of this invention. It is a block diagram which shows the structure of the feature extractor 16 of FIG. It is a graph which shows the vector locus Oc on the complex plane of the complex demodulated signal from the quadrature detector 14 of FIG. (A) is a timing chart showing an observation period To and analysis periods Ta1 to TaN set by the feature quantity pattern calculator 35 of FIG. 2, and (b) is calculated by the feature quantity pattern calculator 35 of FIG. It is a graph which shows the feature-value pattern Fmax and Fmin. It is a block diagram which shows the structure of the intruder identification apparatus 1A which concerns on Embodiment 2 of this invention. It is a block diagram which shows the structure of 16 A of feature extractors which concern on Embodiment 3 of this invention. It is a block diagram which shows the structure of the feature extractor 16B which concerns on Embodiment 4 of this invention.

  Hereinafter, embodiments according to the present invention will be described with reference to the drawings. In addition, in each following embodiment, the same code | symbol is attached | subjected about the same component.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing the configuration of the intruder identification apparatus 1 according to Embodiment 1 of the present invention, and FIG. 2 is a block diagram showing the configuration of the feature extractor 16 in FIG. 3 is a graph showing the vector locus Oc on the complex plane of the complex demodulated signal from the quadrature detector 14 of FIG. 1, and FIG. 4A is set by the feature amount pattern calculator 35 of FIG. FIG. 4B is a graph showing the feature amount patterns Fmax and Fmin calculated by the feature amount pattern calculator 35 of FIG. 2.

  In FIG. 1, the intruder identification device 1 includes a radio transmission circuit 10 including a signal generator 11, a transmission antenna 12, a reception antenna 13, a radio reception circuit 9 including a quadrature detector 14, a memory 15, 20, a feature extractor 16, a discriminator 17, a learning sample data memory 18, a learning unit 19, and a discrimination result output unit 21. In FIG. 2, the feature extractor 16 includes a difference vector calculator 31, a change amount calculator 33, a feature amount pattern calculator 35, and memories 32 and 34. Here, the difference vector calculator 31, the change amount calculator 33, the feature amount pattern calculator 35, the learning unit 19 and the discriminator 17 are specifically implemented by a CPU or a digital signal processor (hereinafter referred to as DSP). Composed. Further, the intruder identification device 1 has a detection area in which the wireless transmission circuit 10 and the wireless reception circuit 9 can wirelessly communicate with each other using the transmission antenna 12 and the reception antenna 13.

As will be described in detail later, the intruder identification apparatus 1 according to the present embodiment generates a predetermined transmission signal and wirelessly transmits the signal using the transmission antenna 12, and the received transmission signal as a reception antenna. 13 wirelessly receives using, in the radio receiving circuit 9 and intruder identification device 1 having the orthogonal detection to demodulate the complex demodulated signal using the transmission signal a transmission signal the radio reception is the generation, respectively In a plurality of analysis periods Ta1 to TaN having a predetermined period length ΔT, a feature amount pattern that changes in each analysis period from the complex demodulated signal and changes differently between intruders and non-intruders. A feature extractor 16 that extracts Fmax and Fmin, and a reference feature amount pattern that represents an intruder entry based on the extracted feature amount patterns Fmax and Fmin. And a discriminator 17 for discriminating whether or not the intruder has intruded using the reference feature amount pattern representing the time when the intruder is not intruding and outputting an identification signal S17 indicating the identification result. Yes.

  In FIG. 1, a wireless transmission circuit 10 outputs a predetermined transmission signal generated by a signal generator 11 to a transmission antenna 12 and radiates it as a radio wave, and outputs the transmission signal to a quadrature detector 14. The radio wave radiated from the transmitting antenna 12 is received by the receiving antenna 13 as a received signal. The quadrature detector 14 uses the transmission signal from the signal generator 11 to quadrature-detect the received signal into the in-phase component that is the real part and the quadrature component that is the imaginary part of the complex demodulated signal, and outputs the result to the memory 15.

  In FIG. 2, the difference vector calculator 31 first samples and discretizes the in-phase component and the quadrature component of the complex demodulated signal stored in the memory 15 at a period of 1/16 second. FIG. 3 shows an example of complex demodulated signals Sc1, Sc2, Sc3, and Sc4 sampled at sampling timings t1, t2, t3, and t4, plotted on a complex plane. As shown in FIG. 3, the complex demodulated signal draws a vector locus Oc on the complex plane. Next, the difference vector calculator 31 calculates the difference between the in-phase component and the quadrature component of the complex demodulated signals Sc1, Sc2,... At each successive two sampling timings, thereby calculating the complex demodulated signal on the complex plane. Difference vectors V12, V23, V34,... Representing the vector locus Oc are calculated and output to the memory 32. For example, in FIG. 3, the in-phase component of the difference vector V12 is calculated by subtracting the in-phase component I1 of the complex demodulated signal Sc1 at timing t2 from the in-phase component I2 of the complex demodulated signal Sc2 at timing t2, and is the quadrature component of the difference vector V12. Is calculated by subtracting the orthogonal component Q1 of the complex demodulated signal Sc1 at timing t2 from the orthogonal component Q2 of the complex demodulated signal Sc2 at timing t2.

  2, the change amount calculator 33 calculates an angle formed by two successive difference vectors based on the difference vectors V12, V23, V34,... Stored in the memory 32. For example, in FIG. 3, the angle θ12 formed by the vector V12 and the vector V23 is calculated, and the angle θ23 formed by the vector V23 and the vector V34 is calculated. Then, the change amount calculator 33 calculates cosine cos θ12, cos θ23,... (Hereinafter referred to as a time series pattern of cos θ) of the calculated angles θ12, θ23,.

  Next, as shown in FIG. 4A, the feature quantity pattern calculator 35 sets an observation period To having a predetermined period length, and a plurality of N each having a period length ΔT within the observation period To. Analysis periods Ta1, Ta2,..., TaN are provided. Here, the length of the observation period To is sufficiently longer than the time period (for example, 1 to 2 seconds) required for the intruder to pass through the detection area of the intruder identification device 1. For example, it has a value within a range of 5 seconds to 10 seconds. Further, the length ΔT of each analysis period Ta1, Ta2,..., TaN is set to a value equal to or shorter than the time period required for the intruder to pass through the detection area of the intruder identification device 1, for example, 2 seconds. Further, the adjacent analysis periods Ta1, Ta2,..., TaN are provided so as to partially overlap each other.

  Further, as shown in FIG. 4B, the feature quantity pattern calculator 35 is based on the time series pattern of cos θ stored in the memory 34, and the time series of cos θ in each analysis period Ta1, Ta2,. Plot the pattern on one graph. Then, the time series pattern of cos θ having the maximum value and the time series pattern of cos θ having the minimum value are output to the discriminator 17 as feature amount patterns Fmax and Fmin, respectively.

  Here, when the radio wave radiated from the transmitting antenna 12 hits the intruder, the radio wave is scattered, so that the vector locus Oc of the complex demodulated signal is smoothly drawn with a circle around the point when no intruder exists. It has the feature of changing. On the other hand, when the transmitting antenna 12 and the receiving antenna 13 are shaken, or when the reflectance of the ground or the like changes due to rain or the like, the vector locus Oc of the complex demodulated signal vibrates in a thin ellipse or a straight line. Have. Furthermore, the change in the vector locus Oc of the complex demodulated signal when the intruder enters is characterized by a short time compared to the change caused by a natural phenomenon. For this reason, as shown in FIG. 4A, an observation period To that is sufficiently longer than the time period required for the intruder to pass through the detection area of the intruder identification apparatus 1 is provided, and the intruder identification apparatus By providing the analysis periods Ta1 to TaN having a period length ΔT that is equal to or shorter than the time period required for the intruder to pass through the one detection area, an analysis period Ta including the period T2 invaded by the intruder, The analysis period Ta including only the period T1 or T3 during which the intruder has not entered can be obtained. Further, the time series pattern of cos θ of the vector locus Oc in the analysis period Ta including the period T2 invaded by the intruder is the time series pattern of cos θ of the vector locus Oc in the analysis period Ta including the period T1 or T3 in which no intruder has entered. It has a large value compared to the series pattern. For this reason, as shown in FIG. 4A, when an intruder enters during the observation period To, the feature amount pattern Fmax calculated by the feature amount pattern calculator 35 is in the period T2 during which the intruder has entered. The feature of the vector locus Oc of the complex demodulated signal is represented, and the feature amount pattern Fmin represents the feature of the vector locus Oc of the complex demodulated signal in the period T1 or T3 when no intruder enters.

  In FIG. 1, a learning sample data memory 18 stores in advance a plurality of learning feature amount patterns of feature amount patterns Fmax and Fmin each having an identification label for identifying whether an intruder is invading or not invading. The learning device 19 performs learning using the learned feature value pattern from the learned feature value pattern memory 18 and the actual measurement values of the feature value patterns Fmax and Fmin from the feature value extractor 16 when the intruder is invading and not intruding. A representative pattern of feature amount patterns Fmax and Fmin at the time of intrusion and non-intrusion by the intruder is created, and a reference feature amount pattern representing the time of intrusion and a reference feature amount pattern representing the time of intrusion of the intruder To the memory 20. The discriminator 17 is based on the feature amount patterns Fmax and Fmin from the feature amount extractor 16, and the reference feature amount pattern indicating the intrusion time stored in the memory 20 and the reference feature indicating the non-intrusion time of the intrusion material. Pattern recognition processing is performed using the quantity pattern / feature quantity extractor 16 to identify whether or not the feature quantity pattern Fmax represents an intruder intrusion, and an identification signal S17 indicating the identification result is generated to output an identification result. To the device 21. The identification result output unit 21 is a speaker and a display device, and outputs a predetermined alarm sound from the speaker and displays a predetermined warning on the display device in response to an identification signal S17 indicating an intruder intrusion.

  Therefore, according to the intruder identification device 1 according to the present embodiment, the feature amount patterns Fmax and Fmin are calculated using the angle θ between two consecutive difference vectors. It is possible to extract a feature that the vector locus Oc of the complex demodulated signal changes smoothly when an intruder enters.

  In addition, an observation period To that is sufficiently longer than the time period required for the intruder to pass through the detection area of the intruder identification apparatus 1 is provided, and the intruder passes through the detection area of the intruder identification apparatus 1. By providing the analysis periods Ta1 to TaN having a period length ΔT that is equal to or shorter than the time period required for the analysis, the feature amount pattern Fmax of the analysis period Ta including the period T2 invaded by the intruder and the intruder has invaded. It is possible to obtain the feature amount pattern Fmin of the analysis period Ta including only the non-period T1 or T3. Since the discriminator 17 performs pattern recognition processing based on the feature amount patterns Fmax and Fmin, the intruder intrusion is distinguished from the fluctuation of the complex demodulated signal caused by the natural phenomenon due to rain or wind, and compared with the prior art. Can be identified with high accuracy. Further, it is possible to extract a feature that the change of the complex demodulated signal due to the intruder intrusion is concentrated in a relatively short time compared to the change of the complex demodulated signal caused by the natural phenomenon.

  Further, since the intruder intrusion is displayed on the display device by the identification result output device 21 and is notified by sound, the monitoring work of the monitoring worker can be facilitated.

  In the above embodiment, the feature quantity pattern calculator 35 extracts a time series pattern of the cosine (cos θ) of the angle θ between two consecutive difference vectors as the feature quantity pattern in each analysis period. However, the present invention is not limited to this, and the feature amount pattern has at least the magnitude and direction of a plurality of difference vectors representing the locus Oc when the complex demodulated signal changes discretely on the complex plane in each analysis period. The time series pattern of one change may be included. Specifically, the feature amount pattern includes an angle between two difference vectors that are continuous or separated by a predetermined time interval, a size of at least one of the two difference vectors, and the angle of the angle. A time-series pattern of a predetermined value calculated based on at least one of a trigonometric function and an absolute value of the trigonometric function of the angle may be used. For example, the predetermined value may be an angle formed by two consecutive difference vectors, a value of a trigonometric function of the angle, an absolute value of a value of the trigonometric function of the angle, or a magnitude of one of the angle and the difference vector. Any one of a value calculated based on the value, a product of the cosine of the angle and the magnitude of one of the difference vectors, and the magnitude of the difference vector may be used. As a result, it is possible to extract feature amount patterns Fmax and Fmin representing a smooth change, which is a feature of the vector locus Oc of the complex demodulated signal when an intruder enters, and to calculate the feature amount patterns Fmax and Fmin with a relatively small amount of calculation. .

Embodiment 2. FIG.
FIG. 5 is a block diagram showing a configuration of an intruder identification apparatus 1A according to Embodiment 2 of the present invention. The intruder identification device 1A includes a radio transmission circuit 10A including a signal generator 11 and a multiplier 41, a transmission array antenna 43, a reception array antenna 46, terminators 44 and 45, a PN code generator 42, A wireless receiver circuit 9A, a plurality of M memories 15-1 to 15-M, a plurality of M feature extractors 16-1 to 16-M, and a plurality of M identifiers 17-1 to 17-M; The learning feature amount pattern memory 18, the learning device 19, the memory 20, the position determination device 22, and the identification result output device 21 are configured. Here, the transmission array antenna 43 is a leaky coaxial cable (LCX (Leaky Coaxial Cable)) having a plurality of L slits each functioning as a plurality of L transmission antennas 43-1 to 43-L and arranged linearly. ). The receiving array antenna is a leaky coaxial cable having a plurality of L slits functioning as a plurality of L receiving antennas 46-1 to 46-L and arranged linearly. Further, the radio reception circuit 9A includes a delay unit 47, a plurality of M multipliers 48-1 to 48-M, and a plurality of M orthogonal detectors 14-1 to 14-M.

  As will be described in detail later, the intruder identification device 1A according to the present embodiment includes a transmission array antenna 43 and a reception array antenna 46 formed of a leaky coaxial cable, and a transmission signal is transmitted using a PN code signal in the wireless transmission circuit 10A. In the wireless communication circuit 9A, the PN code signal is delayed by a plurality of different delay times ΔD1 to ΔDM to generate a plurality of delayed PN code signals, and the received signal is transmitted to the plurality of delayed PN codes. A plurality of despread reception signals are generated by despreading using the signal, and each of the plurality of despread signals is demodulated into a complex demodulated signal and output. The position determiner 22 determines the position of the intruder based on the identification signals S17-1 to S17-M from the identifiers 17-1 to 17-M provided corresponding to the plurality of delay times ΔD1 to ΔDM. Is further provided.

  The PN code generator 42 generates a PN code signal and outputs it to the multiplier 41 and the delay unit 47. The wireless transmission circuit 10 multiplies a predetermined transmission signal generated by the signal generator 11 by a PN code signal, spreads the transmission signal, radiates it as a radio wave using the transmission array antenna 43, and transmits the transmission signal. The signals are output to the quadrature detectors 14-1 to 14-M. The radio waves radiated from the transmission array antenna 43 are received as reception signals by the reception array antenna 46 and output to the multipliers 48-1 to 48 -M. Here, the terminator 44 absorbs radio waves remaining without being radiated by the transmission array antenna 43, and the terminator 45 includes multipliers 48-1 to 48 -M among the radio waves received by the reception array antenna 46. Absorbs radio waves traveling to the opposite side.

  The received signal is received with various delay times depending on a plurality of mutually different propagation paths between the transmitting array antenna 43 and the receiving array antenna 46. In consideration of this, the delay unit 47 delays the PN code signal generated by the PN code generator 42 by a plurality of M delay amounts ΔD1 to ΔDM different from each other, and each delayed PN code signal is delayed. Output to the multipliers 48-1 to 48 -M, respectively. Each of the multipliers 48-1 to 48-M despreads the received signal by multiplying the input received signal by the input delayed PN code signal to generate a despread received signal, and the quadrature detector 14- Output to 1-14-M.

  The orthogonal detectors 14-1 to 14-M, the memories 15-1 to 15-M, the feature extractors 16-1 to 16-M, and the discriminators 17-1 to 17-M are respectively orthogonal according to the first embodiment. Since the detector 14, the memory 15, the feature extractor 16, and the discriminator 17 are configured in the same manner, description of the operation is omitted. Like the discriminator 17, the discriminators 17-1 to 17-M discriminate whether or not the input feature amount pattern Fmax indicates an intruder intrusion, and an identification signal S17-1 indicating the discrimination result. ~ S17-M are generated and output to the position determiner 22. The position determiner 22 is composed of a CPU or a DSP, and based on the delay amount ΔDm corresponding to the identification signal S17-m indicating the intruder intrusion among the identification signals S17-1 to S17-M, the position of the intruder (for example, The distance from the left end of the receiving array antenna 46 to the intruder is determined, and a signal indicating the determination result and the intruder intrusion is output to the identification result output unit 21. In response to this, the identification result output unit 21 outputs a predetermined alarm sound from the speaker and displays a warning indicating the position of the intruder on the display device.

  According to the present embodiment, since the transmission array antenna 43 and the reception array antenna 46 are used as compared with the first embodiment, the detection area of the intruder identification device 1A can be expanded. In particular, since the transmitting array antenna 43 and the receiving array antenna 46 are realized by using a leaky coaxial cable, the detection area can be expanded uniformly in the longitudinal direction of the leaky coaxial cable. Further, complex demodulated signals corresponding to different delay times ΔD1 to ΔDM are generated, the characteristics of the change of the vector locus Oc are extracted for each of the complex demodulated signals corresponding to the delay times ΔD1 to ΔDM, and intruders are intruded. Since the identification is performed, the identification accuracy can be improved and the position of the intruder can be specified as compared with the first embodiment. Furthermore, since the transmission signal is spectrum-spread using the PN code, the resistance to interference waves and interference waves can be improved as compared with the first embodiment.

Embodiment 3 FIG.
FIG. 6 is a block diagram showing a configuration of a feature extractor 16A according to Embodiment 3 of the present invention. The feature extractor 16A includes a frequency converter 51, a memory 52, and a feature amount pattern calculator 35A. The feature extractor 16A according to the present embodiment is characterized in that the feature amount pattern calculator 35A extracts the frequency spectrum pattern of the complex demodulated signal in each analysis period Ta1 to TaN.

  In FIG. 6, the frequency converter 51 and the feature amount pattern calculator 35A are specifically configured by a CPU or a DSP. Similarly to the feature quantity pattern calculator 35 in the first embodiment, the frequency converter 51 sets an observation period To having a predetermined period length, and has a plurality of N each having a period length ΔT within the observation period To. Analysis periods Ta1, Ta2,..., TaN are provided (see FIG. 4A). Then, the complex demodulated signal stored in the memory 15 is Fourier-transformed into a frequency spectrum Z (f) (f is a frequency) and output to the memory 52 for each analysis period Ta1 to TaN. Further, the feature amount pattern calculator 35A plots the frequency spectrum Z (f) in each analysis period Ta1 to TaN stored in the memory 52 in one graph. The pattern of the frequency spectrum Z (f) in the analysis period having the maximum frequency spectrum Z (f) and the pattern of the frequency spectrum Z (f) in the analysis period having the minimum frequency spectrum Z (f) are characterized. The quantity patterns Fmax and Fmin are output to the discriminator 17.

  In general, a complex demodulated signal when an intruder invades is lower than a complex demodulated signal when a natural phenomenon such as fluctuations in reflectance of the ground due to shaking of the transmitting antenna 12 and the receiving antenna 13 or rain falls. Contains a lot of frequency components. For this reason, as shown in FIG. 4A, when an intruder enters during the observation period To, the feature amount pattern Fmax calculated by the feature amount pattern calculator 35A is in the period T2 during which the intruder has entered. The feature of the frequency component of the complex demodulated signal is represented, and the feature amount pattern Fmin represents the feature of the frequency component of the complex demodulated signal in the period T1 or T3 when the intruder has not entered.

  This embodiment has an effect that it is possible to detect an intruder's intrusion by calculating a feature amount pattern representing a frequency spectrum of a complex demodulated signal when the intruder invades.

  In the feature quantity pattern calculator 35A, the frequency spectrum pattern of the complex demodulated signal in each analysis period Ta1 to TaN is extracted. However, the present invention is not limited to this, and the feature quantity pattern including the frequency spectrum pattern is extracted. do it. For example, in the feature quantity pattern calculator 35A, instead of the pattern of the frequency spectrum Z (f), a pattern of the following value Za or Zb may be extracted.

  The value Za represents the magnitude of the rotation component of the vector locus Oc of the complex demodulated signal at each frequency f, and the value Zb represents the magnitude of each frequency component that does not depend on the rotation of the complex demodulated signal. Since any value does not depend on the direction of rotation of the vector locus Oc of the complex demodulated signal, the data amount of the learning feature amount pattern can be reduced.

  In the present embodiment, the frequency converter 51 Fourier-transforms the complex demodulated signal into the frequency spectrum. However, the present invention is not limited to this, and the complex demodulated signal may be wavelet transformed into the frequency spectrum. By using the Fourier transform or the wavelet transform, the complex demodulated signal can be frequency-converted relatively easily.

Embodiment 4 FIG.
FIG. 7 is a block diagram showing a configuration of a feature extractor 16B according to Embodiment 4 of the present invention. The feature extractor 16B includes a matrix calculator 61, a memory 62, and a feature amount pattern calculator 35B. The feature extractor 16B according to the present embodiment is characterized in that the feature amount pattern calculator 35B has extracted a time-series pattern of changes in the variance of the covariance matrix of the complex demodulated signal in each analysis period Ta1 to TaN.

  In FIG. 7, the matrix calculator 61 and the feature amount pattern calculator 35B are specifically configured by a CPU or DSP. The matrix calculator 61 calculates a covariance matrix from the in-phase component and the quadrature component of the complex demodulated signal stored in the memory 15 for each sub-observation period having a predetermined period length shorter than the period length ΔT of the analysis period, The variance of the covariance matrix is calculated and output to the memory 62. Similar to the feature value pattern calculator 35 in the first embodiment, the feature value pattern calculator 35B sets an observation period To having a predetermined length, and has a period length ΔT within the observation period To. A plurality of N analysis periods Ta1, Ta2,..., TaN are provided (see FIG. 4A). Then, using the variance stored in the memory 62, the time series pattern of variance is plotted in one graph for each analysis period Ta1 to TaN. Then, the variance time series pattern of the analysis period having the maximum variance and the variance time series pattern of the analysis duration having the minimum variance are output to the discriminator 17 as the feature amount patterns Fmax and Fmin.

  As described above, the vector locus Oc of the complex demodulated signal at the time of intrusion by the intruder has a characteristic of smoothly changing in a circle. For this reason, the variance of the complex demodulated signal at the time of intrusion by the intruder becomes larger than the variance of the complex demodulated signal at the time of non-intrusion. Therefore, as shown in FIG. 4A, when an intruder enters during the observation period To, the feature amount pattern Fmax calculated by the feature amount pattern calculator 35A is a complex of the period T2 during which the intruder has entered. The feature of the vector locus Oc of the demodulated signal is represented, and the feature amount pattern Fmin represents the feature of the vector locus Oc of the complex demodulated signal in the period T1 or T3 when no intruder has entered.

  This embodiment has an effect that the feature amount representing the vector locus Oc of the complex demodulated signal at the time of intruder intrusion can be calculated to detect the intruder intrusion.

  In the present embodiment, the feature amount pattern calculator 35B extracts the time series pattern of the variance change of the covariance matrix of the complex demodulated signal in each analysis period Ta1 to TaN, but the present invention is not limited to this. What is necessary is just to extract the time series pattern of the change of the variance of the covariance matrix or the change of the eigenvalue. For example, the feature amount pattern calculator 35B may extract a time series pattern of eigenvalues of the covariance matrix of the complex demodulated signal. Here, the first principal component and the second principal component are referred to in descending order of the eigenvalues.

  Further, in the feature amount pattern calculator 35B, a change in the length of the locus when the complex demodulated signal changes discretely on the complex plane in each analysis period Ta1 to TaN, or a change in the area of the region surrounded by the locus. These time series patterns may be extracted. For example, the vector locus Oc length (orbit length) Lo and the area (orbit area) Ao of the region surrounded by the vector locus Oc of the complex demodulated signal are the in-phase components I (i) and Q (i) ( i is the sampling timing within the sub-analysis period.)

  For the trajectory length Lo and the trajectory area Ao, the sum is calculated for all sampling timings i within the sub-analysis period. The trajectory length Lo and trajectory area Ao of the complex demodulated signal at the time of intrusion by the intruder has a characteristic that it has a larger value than the trajectory length Lo and trajectory area Ao of the complex demodulated signal at the time of non-intrusion.

  Note that in the intruder identification device 1A according to the second embodiment, instead of the feature extractors 16-1 to 16-M, the feature extractor 16A according to the third embodiment or the feature extractor 16B according to the fourth embodiment. May be used.

  In each of the above embodiments, the feature quantity patterns extracted by the feature extractors 16 and 16-1 to 16-M are: (1) The complex demodulated signal is discretely distributed on the complex plane in each analysis period Ta1 to TaN. A feature amount pattern including a time series pattern of a change in at least one of the magnitude and direction of a plurality of vectors representing a trajectory when changing, and (2) a frequency spectrum pattern of a complex demodulated signal in each analysis period Ta1 to TaN. A feature amount pattern including (3) a feature amount pattern including a variation in the covariance matrix of the complex demodulated signal in each analysis period Ta1 to TaN, or a time series pattern of a change in eigenvalue, or (4) each analysis period Ta1 to Change in length of trajectory when complex demodulated signal changes discretely on complex plane in TaN, or region surrounded by the trajectory It was characteristic quantity pattern including a time-series pattern of change in the area. However, the present invention is not limited to this, and the feature extractors 16 and 16-1 to 16-M change from the complex demodulated signal in each analysis period Ta1 to TaN and change in each analysis period Ta1 to TaN and intruders. What is necessary is just to extract the feature-value pattern which changes so that it may differ at the time of an intrusion, and the time of non-intrusion.

  Further, in each of the above embodiments, the discriminators 17, 17-1 to 17-M use the reference feature amount pattern that represents the time when the intruder enters and the reference feature amount pattern that represents the time when the intruder does not enter. It is sufficient to use at least one of a reference feature amount pattern representing the time when the intruder enters and a reference feature amount pattern representing the time when the intruder does not enter.

  Furthermore, in each of the above embodiments, the intruder identification devices 1 and 1A have identified that an intruder has entered the detection area. However, the present invention is not limited to this, and the intruder is a small animal such as a dog or a cat. May be identified as having entered the detection area.

  As described above in detail, the intruder identification device according to the present invention changes from a complex demodulated signal in each analysis period in a plurality of analysis periods each having a predetermined period length, and is different from that at the time of intrusion intrusion. Feature extraction means for extracting a feature amount pattern that changes differently at the time of intrusion, a reference feature amount pattern that represents the time of intrusion, based on the extracted feature amount pattern, and a time when the intruder is not intruding And an identification means for identifying whether or not the intruder has intruded using at least one of the reference feature amount patterns representing the output and outputting an identification signal indicating the identification result. The characteristic amount pattern peculiar to the complex demodulated signal is extracted, and it is possible to accurately identify that the intruder has intruded as compared with the prior art.

DESCRIPTION OF SYMBOLS 1,1A Intruder identification device, 9,9A radio receiving circuit, 10,10A radio transmitting circuit, 11 signal generator, 12 transmitting antenna, 13 receiving antenna, 14, 14-1 to 14-M quadrature detector, 15, 15-1 to 15-M, 20, 32, 52, 62 Memory, 16 Feature extractor, 17, 17-1 to 17-M Discriminator, 18 Learning feature value pattern memory, 19 Learner, 21 Discrimination result output device , 22 position determiner, 31 vector calculator, 33 fluctuation amount calculator, 35, 35A, 35B feature amount pattern calculator, 41, 48-1 to 48-M multiplier, 42 PN code generator, 43 transmission array antenna , 44, 45 terminator, 46 receiving array antenna, 47 delay unit, 51 frequency converter.

Claims (10)

Transmitting means for generating a predetermined transmission signal and wirelessly transmitting using a transmission antenna;
Wirelessly received using the receiving antennas a transmission signal transmitted above, and a receiving means for demodulating the complex demodulated signal received signals the wireless receiver to quadrature detection by using the transmission signal the occurrence intrusion In the object identification device,
In a plurality of analysis periods each having a predetermined period length, feature amount patterns that change also in each analysis period and change differently at the time of intrusion and non-intrusion are extracted from the complex demodulated signal. Feature extraction means;
Based on the extracted feature amount pattern, the intruder has intruded using at least one of the reference feature amount pattern indicating the intruder entry time and the reference feature amount pattern indicating the intruder non-intrusion time. An intruder identification apparatus comprising: an identification unit that identifies whether or not the identification result is output, and outputs an identification signal indicating the identification result.   The feature pattern includes a time-series pattern of a change in at least one of a magnitude and a direction of a plurality of vectors representing a locus when the complex demodulated signal changes discretely on a complex plane in each analysis period. The intruder identification apparatus according to claim 1, wherein:   The intruder identifying apparatus according to claim 1, wherein the feature amount pattern includes a frequency spectrum pattern of a complex demodulated signal in each analysis period.   4. The intruder identification apparatus according to claim 3, wherein the feature extraction means performs Fourier transform or wavelet transform on the complex demodulated signal into the frequency spectrum in each analysis period.   2. The intruder identification apparatus according to claim 1, wherein the feature amount pattern includes a time-series pattern of a change in covariance matrix of the complex demodulated signal or a change in eigenvalue in each analysis period.   The feature amount pattern is a time-series pattern of a change in the length of the locus when the complex demodulated signal changes discretely on the complex plane in each analysis period, or a change in the area of the area surrounded by the locus. The intruder identification device according to claim 1, further comprising:   The intruder identification apparatus according to claim 1, wherein each of the transmission antenna and the reception antenna is an array antenna.   8. The intruder identification device according to claim 7, wherein each of the array antennas is a leaky coaxial cable.   9. The intruder identification apparatus according to claim 1, wherein the transmission means transmits the generated transmission signal by performing spectrum spread using a PN code signal. The receiving means delays the PN code signal by a plurality of different delay times to generate a plurality of delayed PN code signals, despreads the received signal using the plurality of delayed PN code signals, Generate a despread received signal, each of the plurality of despread signals is quadrature detected using the generated transmission signal, demodulated into a complex demodulated signal, and output,
The intruder identification device is
A plurality of the feature extraction means respectively provided corresponding to the plurality of delay times;
A plurality of the identifying means respectively provided corresponding to the plurality of delay times;
10. The intruder identification apparatus according to claim 9, further comprising position determining means for determining a position of the intruder based on an identification signal from each identification means.

Images