JP7846571B2 - Ground-based sensor detection device - Google Patents

Description

Applicable under Article 30, Paragraph 2 of the Patent Act. Published on the website on November 1, 2021. Collection of papers on CD published on November 1, 2021.

This invention relates to a ground beacon detection device for detecting ground beacons having a resonant frequency at one of a plurality of predetermined frequencies from a moving railway vehicle.

For example, in automatic train stop (ATS) systems used to improve railway safety, frequency-variable type systems have traditionally been employed to detect ground beacons installed along the tracks. In frequency-variable ATS systems, an oscillator circuit, consisting of a train-mounted beacon (a coil with loosely coupled double windings) forming a feedback loop, constantly oscillates. Electromagnetic coupling occurs when the ground beacon (coil + capacitor) and the train-mounted beacon are in close proximity. At this point, the oscillation frequency of the train-mounted beacon changes to a resonant frequency determined by the ground beacon (frequency variation), which is then used to generate information such as a stop indication. However, this frequency-variable method relies heavily on analog circuits, leading to difficulties in component supply and design reproducibility, and alternative methods have been sought.

Therefore, the device described in Patent Document 1, in order to achieve digitalization, transmits a spectrum signal from the on-board unit and detects the resonant frequency of the ground unit by performing FFT analysis on the frequency reflected from the ground unit. The device described in Patent Document 2 employs a method similar to that of Patent Document 1, improving the detection accuracy of ground units while suppressing power consumption. Furthermore, the device described in Patent Document 3 uses a method of repeatedly transmitting a swept transmission wave near the resonant frequency of the ground unit at a constant period and receiving the reflection. The device described in Patent Document 4 performs detection with enhanced noise immunity by calculating complex correlation values of ground unit frequencies at lower and higher frequencies in addition to the target ground unit frequency.

Patent No. 4596819 Patent No. 5951240 Patent No. 3959015 Patent No. 4879219

Here, the inventions described in Patent Documents 1 to 3 above all focus only on the frequency of the ground beacon, and therefore have reliability issues as they erroneously detect noise at the same frequency as the ground beacon. In contrast, the invention described in Patent Document 4 has improved noise immunity, but it has the problem of having a complex detection algorithm.
This invention was made in view of the above-mentioned problems, and its purpose is to improve the reliability of ground signal detection while using a simple algorithm.

(Aspects of the invention)
The following embodiments of the invention are illustrative of the configuration of the present invention and are described in separate sections to facilitate understanding of the diverse configurations of the present invention. Each section does not limit the technical scope of the present invention, and while taking into consideration the best mode for carrying out the invention, the technical scope of the present invention may also include modifications to some of the components of each section, such as substitution, deletion, or addition of other components.

(1) A ground beacon detection device for detecting a ground beacon installed along a railway track, having a resonant frequency of one of a plurality of predetermined frequencies, from a moving railway vehicle, comprising: a primary coil and a secondary coil that are loosely coupled to each other, and an on-board unit that electromagnetically couples with the ground beacon when in close proximity to the ground beacon; a transmission processing unit that injects a transmission wave of a discrete frequency spectrum containing a plurality of frequencies encompassing all of the plurality of predetermined frequencies at a frequency interval smaller than the frequency interval of the plurality of predetermined frequencies into the primary coil; a frequency discrimination unit that extracts the discrete frequency spectrum from a received wave received by the secondary coil; a spectral feature extraction unit that extracts a predetermined shape from a spectral waveform defined by the levels of each frequency component of the discrete frequency spectrum extracted by the frequency discrimination unit; and a ground beacon detection unit that detects the ground beacon according to the presence or absence of the predetermined shape.

The ground beacon detection device described in this section detects ground beacons from a moving railway vehicle and includes a vehicle-mounted unit, a transmission processing unit, a frequency discrimination unit, a spectral feature extraction unit, and a ground beacon detection unit. The vehicle-mounted unit has a primary coil and a secondary coil that are loosely coupled to each other, and constantly transmits and receives signals between them. When the railway vehicle is in motion, if the vehicle-mounted unit approaches a ground beacon near the tracks, the ground beacon and the vehicle-mounted unit become electromagnetically coupled. The transmission processing unit injects a transmission wave having a discrete frequency spectrum into the primary coil of the vehicle-mounted unit. The discrete frequency spectrum of this transmission wave includes multiple frequencies that encompass all of the multiple predetermined frequencies that the ground beacon to be detected may have as its resonant frequency, at frequency intervals smaller than the multiple predetermined frequencies.

The transmission wave, as described above, injected into the primary coil is received by a secondary coil that is loosely coupled to the primary coil. The frequency discrimination unit extracts a discrete frequency spectrum from the received wave received by the secondary coil, containing the same frequency components as those generated by the transmission processing unit. The spectral feature extraction unit then generates a spectral waveform from the discrete frequency spectrum extracted by the frequency discrimination unit, defined by the levels of each frequency component, and further extracts a predetermined shape from this spectral waveform. Finally, the ground coil detection unit detects ground coils based on the presence or absence of the predetermined shape extracted from the spectral waveform.

Here, when the ground coil and the on-board coil electromagnetically couple as described above, one of the multiple predetermined frequencies that each on-board coil holds as information is transmitted to the on-board coil as the resonant frequency. That is, the resonant frequency affects the received wave received by the on-board coil's secondary coil, and a shape appears in the spectral waveform generated based on the received wave where the level of the frequency component at the same frequency as the ground coil's resonant frequency is increased. Therefore, by setting a predetermined shape that increases in level at the resonant frequency as the shape extracted from the spectral waveform by the spectral feature extraction unit, it is detected that the on-board coil has electromagnetically coupled to a ground coil having that resonant frequency. As a result, since the resonant frequency appearing in the spectral waveform is extracted by the waveform shape rather than the magnitude of the level, false detection of simply high-level noise and the influence of absolute level fluctuations due to various factors are suppressed, thereby increasing the reliability of ground coil detection. Moreover, the extraction of the predetermined shape from the spectral waveform is achieved using a relatively simple algorithm through measures such as simplifying the processing data.

(2) In the above item (1), the transmission processing unit injects the transmitted wave into the primary coil at a constant voltage, and the frequency discrimination unit performs level conversion on the extracted discrete frequency spectrum based on the received voltage of each frequency component.
The ground-based coil detection device described in this section injects the transmitted wave into the primary coil of the on-board coil at a constant voltage when the transmission processing unit injects the transmitted wave. The frequency discrimination unit then performs level conversion on the discrete frequency spectrum extracted from the received wave of the secondary coil based on the received voltage of each frequency component. Therefore, the spectral feature extraction unit extracts a predetermined shape from the spectral waveform generated based on the discrete frequency spectrum level-converted by the frequency discrimination unit. As a result, voltage differences between frequency components that were not present in the transmitted wave appear in the received wave due to electromagnetic coupling with the ground coil, and these voltage differences are further increased by the predetermined level conversion operation. Consequently, it becomes easier to extract a predetermined shape from the spectral waveform in which such voltage differences are reflected in the level, thereby improving the detection accuracy of the ground coil.

(3) In item (1) above, the ground coil detection device wherein the discrete frequency spectrum has a bandwidth that includes the anti-resonance frequency of the combined impedance when the on-board coil and the ground coil are electromagnetically coupled.
The ground coil detection device described in this section has a bandwidth that includes the anti-resonant frequency of the combined impedance during electromagnetic coupling between the on-board coil and the ground coil, in the discrete frequency spectrum of the transmitted wave generated by the transmission processing unit and the discrete frequency spectrum extracted by the frequency discrimination unit. Since the anti-resonant frequency appears as a lower level than the resonant frequency during electromagnetic coupling between the on-board coil and the ground coil, the bandwidth of the discrete frequency spectrum is set so that the above-mentioned region below the resonant frequency (predetermined frequency) is covered for all of the multiple predetermined frequencies that the ground coil may have. Furthermore, by including the shape showing the effect of the anti-resonant frequency during electromagnetic coupling in the predetermined shape extracted by the spectral feature extraction unit, the distinction between the ground coil and noise becomes clearer, and the reliability of ground coil detection is further enhanced.

(4) In the above item (3), the spectral feature extraction unit extracts a predetermined shape in which the level rises from the anti-resonance frequency toward the high-frequency side to the resonant frequency at the time of electromagnetic coupling between the on-board unit and the ground unit, and then the level falls toward the high-frequency side.
The ground coil detection device described in this section is characterized in that the predetermined shape extracted from the spectral waveform by the spectral feature extraction unit is more specifically identified as follows: The predetermined shape is set to one in which the level rises from a relatively low anti-resonant frequency towards the high-frequency side to the resonant frequency, and then the level decreases towards the high-frequency side. As a result, a characteristic shape that appears when electromagnetically coupled with a ground coil is set, in which the level rises significantly from the anti-resonant frequency to the resonant frequency and then decreases, making it easier to distinguish from noise and allowing ground coils to be detected with greater accuracy.

(5) In the above item (1), the spectral feature extraction unit approximates the spectral waveform in a piecewise manner by combining any line segments from three types of line segments: positive gradient line segments whose level increases toward the higher frequency side, negative gradient line segments whose level decreases toward the higher frequency side, and flat line segments that do not fall under either the positive gradient line segments or the negative gradient line segments.
The ground-based signal detection device described in this section uses a spectral feature extraction unit to approximate the spectral waveform, defined by the levels of the frequency components of the discrete frequency spectrum, into a piecewise linear shape by combining arbitrary line segments from three types of line segments. These three types of line segments are positive gradient line segments whose level increases towards higher frequencies, negative gradient line segments whose level decreases towards higher frequencies, and flat line segments that do not fall under either positive or negative gradient line segments. As a result, the amount of data processed in subsequent stages is compressed compared to when the spectral waveform is handled directly, making it easier to implement programs for subsequent processing.

(6) In the above item (5), the spectral feature extraction unit calculates, for each line segment of the polylinearly approximated spectral waveform, at least the frequency and level of the starting point which is the endpoint on the low-frequency side and the frequency and level of the ending point which is the endpoint on the high-frequency side as a line segment primary attribute group.
The ground-based signal detector described in this section uses a spectral feature extraction unit to calculate a primary attribute group for each line segment of the spectral waveform, which is approximated as a piecewise linear waveform as described in section (5) above. This primary attribute group includes at least the frequency and level of the starting point, which is the low-frequency endpoint of each line segment, and the frequency and level of the ending point, which is the high-frequency endpoint of each line segment. This allows the characteristics of each line segment to be represented as a data set that is easier to process.

(7) In the above item (6), the spectral feature extraction unit calculates, for each line segment of the polylinearly approximated spectral waveform, at least the level difference between the starting point and the ending point, and the bandwidth which is the frequency difference between the starting point and the ending point, as a group of quadratic line segment attributes.
The ground-based signal detector described in this section uses a spectral feature extraction unit to calculate a secondary line segment attribute group for each line segment of the spectral waveform, which is approximated as a piecewise linear waveform as described in section (5) above, using the primary line segment attribute group described in section (6) above. This secondary line segment attribute group includes at least the level difference between the start and end points of each line segment and the bandwidth, which is the frequency difference between the start and end points of each line segment. As a result, the characteristics of each line segment are represented as a data set that is easier to process.

(8) In the above item (7), the spectral feature extraction unit extracts a peak shape from the spectral waveform approximated as a piecewise linear shape, which changes from the positive gradient line segment to the negative gradient line segment toward the higher frequency side, and/or changes from the positive gradient line segment to the negative gradient line segment via the flat line segment of a small bandwidth toward the higher frequency side.
The ground-based signal detector described in this section uses a spectral feature extraction unit to extract peak shapes from a spectral waveform that has been approximated as a piecewise linear shape, as described in section (5) above. These peak shapes are the portion that changes from a positive gradient line segment to a negative gradient line segment towards the higher frequency side, and/or the portion that changes from a positive gradient line segment to a negative gradient line segment via a flat line segment with a small bandwidth towards the higher frequency side.

Here, a flat line segment with a small bandwidth refers to a flat line segment with a relatively small bandwidth. Whether or not a line segment has a small bandwidth is determined by considering the bandwidth of the preceding and succeeding positive and negative gradient line segments, as well as experimental results. By using such peak shapes for processing, the amount of data processed in subsequent stages is compressed compared to when the entire spectral waveform approximated as a piecewise linear waveform is handled. This simplifies the algorithms used for processing in the spectral feature extraction and ground signal detection units, thereby reducing the CPU load. The number of peak shapes extracted from the spectral waveform may be zero if there are no peak shapes, or multiple if there are multiple peak shapes.

(9) In the above item (8), the spectral feature extraction unit, when the positive gradient line segments constituting the peak shape are treated as peak positive gradient line segments and the negative gradient line segments constituting the peak shape are treated as peak negative gradient line segments, calculates for each of the peak shapes, at least the normalized peak level, which is the sum of the absolute value of the level difference of the peak positive gradient line segment and the absolute value of the level difference of the peak negative gradient line segment; the normalized peak bandwidth, which is the sum of the bandwidth of the peak positive gradient line segment and the bandwidth of the peak negative gradient line segment; and the normalized peak frequency, which is the average of the frequency of the endpoint of the peak positive gradient line segment and the frequency of the starting point of the peak negative gradient line segment, as a normalized peak primary attribute group. The ground coil detection unit then uses the normalized peak primary attribute group to determine whether or not the peak shape represents a ground coil.

The ground-based signal detector described in this section calculates the following normalized peak primary attribute group for each peak shape extracted by the spectral feature extraction unit as described in section (8) above, by treating the positive gradient line segments constituting the peak shape as peak positive gradient line segments and the negative gradient line segments constituting the peak shape as peak negative gradient line segments. Specifically, the normalized peak primary attribute group includes at least the normalized peak level, normalized peak bandwidth, and normalized peak frequency. The normalized peak level is the sum of the absolute values of the level differences between the peak positive gradient line segments and the peak negative gradient line segments. The normalized peak bandwidth is the sum of the bandwidths of the peak positive gradient line segments and the peak negative gradient line segments, and if the peak shape includes flat line segments with small bandwidths, these are also taken into account in the calculation. Furthermore, the normalized peak frequency is the average of the frequency at the end point of the peak positive gradient line segment and the frequency at the beginning point of the peak negative gradient line segment.

The ground coil detection unit then uses the normalized primary peak attribute group of each peak shape described above to determine whether or not the peak shape represents a ground coil, that is, whether or not it is a peak shape that appears due to electromagnetic coupling between the ground coil and the on-board coil. By utilizing this normalized primary peak attribute group, characteristic peak shapes, including the resonant and anti-resonant frequencies, which change constantly depending on the distance between the on-board and ground coils, can be easily extracted using software. Furthermore, since all the attributes included in the normalized primary peak attribute group are simple indicators, the software processing for determining the peak shape representing a ground coil is simple and can be handled by a low-speed CPU.

Furthermore, the normalized peak level, one of the primary attribute groups of the normalized peak, is calculated by integrating the positive and negative gradient line segments. This results in a significantly higher signal-to-noise ratio compared to conventional peak levels, which rely solely on maximum values detected from simple spectral analysis such as FFT. This enables a highly reliable ground tone detection method with superior noise immunity. Additionally, the power consumption of the transmitting amplifier used in the transmission processing unit is reduced compared to conventional methods, contributing to miniaturization and cost reduction of the equipment. Moreover, because the normalized peak level is a relative level that integrates the level differences of the positive and negative gradient line segments, the absolute levels cancel each other out, eliminating the influence of circuit gain, on-board tone coupling, and other factors that cause absolute level fluctuations. Therefore, the reliability of ground tone detection is further improved.

(10) In the above item (9), the spectral feature extraction unit calculates a normalization gradient, which is the ratio of the normalized peak level to the normalized peak bandwidth, for each of the peak shapes as a group of normalized peak secondary attributes, and the ground coil detection unit uses the normalization gradient to determine whether or not the peak shape indicates a ground coil.
The ground tonearm detection device described in this section has a spectral feature extraction unit that, for each of the peak shapes extracted as described in section (8) above, further calculates a normalized peak secondary attribute group. This normalized peak secondary attribute group includes a normalization gradient, which is the ratio of the normalized peak level to the normalized peak bandwidth. The ground tonearm detection unit then uses the normalization gradient of each peak shape calculated by the spectral feature extraction unit to determine whether or not that peak shape represents a ground tonearm.

In other words, the normalization gradient, which is the ratio of the normalized peak level to the normalized peak bandwidth, tends to be smaller (gentler slope) for ground benton coupling than for single-frequency noise when calculated, for example, by dividing the normalized peak level by the normalized peak bandwidth. Therefore, even if the noise frequency and ground benton frequency are the same, the above-mentioned relationship in the magnitude of the normalization gradients prevents single-frequency noise from being mistakenly detected as a ground benton, thus allowing for accurate detection of the ground benton.

(11) In the above item (10), the spectral feature extraction unit further calculates a normalized gradient level distribution ratio for each of the peak shapes, which is the ratio of the absolute value of the level difference of the peak positive gradient line segment to the absolute value of the level difference of the peak negative gradient line segment, as the normalized peak secondary attribute group, and the ground beacon detection unit determines whether or not the peak shape indicates a ground beacon using the normalized gradient level distribution ratio.
The ground-based transponder detection device described in this section has a spectral feature extraction unit that calculates a normalized gradient level distribution ratio for each peak shape as a normalized peak secondary attribute group. This normalized gradient level distribution ratio is the ratio of the absolute value of the level difference of the peak positive gradient line segment to the absolute value of the level difference of the peak negative gradient line segment. The ground-based transponder detection unit then uses the normalized gradient level distribution ratio of each peak shape calculated by the spectral feature extraction unit to determine whether or not that peak shape represents a ground-based transponder.

Here, during electromagnetic coupling between the on-board and ground coils, the influence of magnetic flux causes side lobes that should not be detected to be generated before and after the main lobe that should be detected. While the main lobe exhibits a peak shape where the anti-resonant frequency is lower than the resonant frequency, the side lobes exhibit a peak shape where the anti-resonant frequency is higher than the resonant frequency. That is, in the main lobe's peak shape, the absolute value of the level difference of the positive gradient line segment from the anti-resonant frequency to the resonant frequency is greater than the absolute value of the level difference of the negative gradient line segment. In contrast, in the side lobe's peak shape, the absolute value of the level difference of the negative gradient line segment from the resonant frequency to the anti-resonant frequency is greater than the absolute value of the level difference of the positive gradient line segment. Therefore, a normalized gradient level distribution ratio is used to compare the magnitudes of the absolute values of the level differences of the peak positive gradient line segment and the peak negative gradient line segment, thereby stably suppressing the detection of unnecessary side lobes.

(12) In item (8) above, the ground beacon detection unit is a ground beacon detection device that determines the presence or absence of the ground beacon according to the number of peak shapes extracted simultaneously from the spectral waveform which is approximated as a piecewise linear shape.
The ground beacon detection device described in this section determines the presence or absence of a ground beacon based on the number of peak shapes simultaneously extracted from the spectral waveform approximated as a piecewise linear waveform by the spectral feature extraction unit. Here, noise from vehicle equipment and track circuit equipment generally consists mostly of pulse noise originating from the power supply system, and although it is short-lived, it generates broadband components, so in such cases multiple peak shapes are extracted simultaneously. Therefore, by using the number of peak shapes for ground beacon detection, noise can be eliminated more efficiently, thereby reducing the CPU processing load while improving the accuracy of ground beacon detection.

(13) In the above item (11), the ground beacon detection unit estimates the occurrence of equipment malfunction and environmental abnormality in the ground beacon detection device according to the number of peak shapes simultaneously extracted from the spectral waveform approximated as a piecewise linear shape, and the number of peak shapes counted as candidates for the ground beacon, determined from the simultaneously extracted peak shapes using the normalized peak primary attribute group and the normalized peak secondary attribute group.

The ground-based transponder detection device described in this section estimates equipment and environmental abnormalities based on the number of peak shapes simultaneously extracted from the spectral waveform approximated as a piecewise linear shape by the spectral feature extraction unit, and the number of peak shapes counted as candidate transponders from among those simultaneously extracted peak shapes. The peak shapes counted as candidate transponders are determined using the normalized peak primary attribute group and the normalized peak secondary attribute group. That is, if multiple peak shapes are simultaneously extracted from the spectral waveform, it is highly likely that this is due to noise, as described in section (12) above, and this is used to estimate environmental abnormalities such as noise. Furthermore, if multiple peak shapes are counted as candidate transponders, it is highly likely that this indicates an equipment malfunction in the transponder detection device, and this is used to estimate equipment abnormalities such as failures. This enables high-precision detection of transponders, as well as equipment monitoring and noise monitoring.

(14) A ground beacon detection method for detecting a ground beacon installed along a railway track, having a resonant frequency of one of a plurality of predetermined frequencies, from a moving railway vehicle, comprising: installing a vehicle-mounted beacon on the railway vehicle having a primary coil and a secondary coil that are loosely coupled to each other, and which electromagnetically couples with the ground beacon when it is close to the ground beacon; injecting a transmission wave of a discrete frequency spectrum containing a plurality of frequencies that encompass all of the plurality of predetermined frequencies at a frequency interval smaller than the frequency interval of the plurality of predetermined frequencies into the primary coil; extracting the discrete frequency spectrum from the received wave received by the secondary coil; extracting a predetermined shape from the spectral waveform defined by the level of each frequency component of the extracted discrete frequency spectrum; and detecting the ground beacon according to the presence or absence of the predetermined shape.
The ground beacon detection method described in this section is performed by the ground beacon detection device described in item (1) above, and thus performs the same function as the ground beacon detection device described in item (1) above.

Because of the above-described configuration, this invention makes it possible to improve the reliability of ground signal detection while using a simple algorithm.

This is a schematic diagram illustrating the configuration of a ground signal detection device according to an embodiment of the present invention. This is a functional block diagram illustrating the configuration and functions of a ground signal detection device according to an embodiment of the present invention. This is an illustrative diagram of a spectral waveform to illustrate the basic concept of a ground transponder detection method using a ground transponder detection device according to an embodiment of the present invention. An example of a discrete frequency spectrum is shown. This shows how the spectral waveform is approximated as a piecewise linear structure. This is an illustrative diagram showing how the peak shape is normalized. This flowchart shows an example of a method for determining the peak shape of a potential ground tonearm. This flowchart shows an example of a final determination method for ground transponders. This is an illustrative diagram illustrating a method for distinguishing between ground signals and noise using a normalized gradient. This is an illustrative diagram illustrating how to distinguish between the main lobe and the side lobes using the normalized gradient level distribution ratio. This is an illustrative diagram illustrating a method for distinguishing between ground signals and noise using the number of peak shapes. This shows how the absolute level of the spectral waveform fluctuates.

Hereinafter, embodiments for carrying out the present invention will be described based on the attached drawings. Here, detailed descriptions of parts identical to or corresponding to those in the prior art will be omitted, and throughout the drawings, identical or corresponding parts are indicated by the same reference numerals.
Figure 1 schematically shows an example of the configuration of a ground beacon detection device 10 according to an embodiment of the present invention, with the ground beacon detection device 10 being used in an automatic train stop (ATS) system as an example. As shown in the figure, the ground beacon detection device 10 includes an on-board device 20 and an on-board beacon 12 installed on a railway vehicle 86, and detects a ground beacon 80 installed near the tracks 84 when the on-board beacon 12 approaches the ground beacon 80 as the railway vehicle 86 is moving.

The on-board unit 20 includes a transmission processing unit 22, a reception processing unit 32, and a control unit 70. The on-board unit 12 consists of a primary coil 14 and a secondary coil 16 that are loosely coupled to each other, and transmission and reception constantly occur between them. The same type as that used in a frequency-variable ATS may be used. The primary coil 14 and secondary coil 16 are not limited to these, but they are standardized to have the same cross-sectional area and number of turns. As will be described in detail later, the transmission wave generated by the transmission processing unit 22 is injected into the primary coil 14 of the on-board unit 12, and a voltage reflecting the electromagnetic interaction with the surroundings of the on-board unit 12 appears in the secondary coil 16. This voltage is then taken in as a received wave by the reception processing unit 32. Note that the names "primary coil" and "secondary coil" are for convenience, and their usage may be reversed from the above.

The receiving processing unit 32 identifies the frequency of the ground beacon 80, which is electromagnetically coupled to the on-board beacon 12. The control unit 70 is responsible for the overall control of the on-board device 20. Based on the frequency of the ground beacon 80 identified by the receiving processing unit 32, it sends out train control signals and receives speed information of the railway vehicle 86. The control unit 70 may also perform speed limit checks based on the received speed information. The vehicle control unit 78 is not part of the ground beacon detection device 10, but controls the movement of the railway vehicle 86. Based on the train control signals sent from the control unit 70, it performs braking, speed control, etc., of the railway vehicle 86.

On the other hand, the ground beacon 80, which the ground beacon detection device 10 detects, has the same structure as those used in frequency-variable ATS systems. It uses a loop coil as an inductor, resonating in series with a capacitor connected in series with the loop coil, and possesses the resonant frequency itself as information. The capacitor value is selected so that the resonant frequency matches one of several predetermined frequencies set in advance as a specification. Furthermore, in the ground beacon 80, two or more of the multiple predetermined frequencies may be set as switchable resonant frequencies. For each of these multiple predetermined frequencies (for example, f1 to f8 shown at the top of Figure 3), various information such as stop indications and speed indications are defined. The definitions, the number of predetermined frequencies used, and the frequency values of the predetermined frequencies vary depending on the railway operator.

A relay 82 is inserted between the inductor (loop coil) and capacitor of the ground beacon 80. This relay 82 is normally open, and when the above-mentioned information is to be notified to the railway vehicle 86, the relay 82 operates, forming contacts and activating the LC resonant circuit. The operation of the relay 82 is controlled by an external control input to the ground beacon 80, indicating the presence of a train ahead. For example, if there is no train ahead and a brake command is not required, the coil of the ground beacon 80 is opened by the relay 82, and electromagnetic coupling does not occur even if the on-board beacon 12 is close to the ground beacon 80; therefore, a stop indication is not recognized on board the vehicle.

Next, Figure 2 shows a functional block diagram of the ground beacon detection device 10 according to an embodiment of the present invention. The specific ground beacon detection method using the ground beacon detection device 10 will be explained in accordance with this functional block diagram. The transmission processing unit 22 and on-board beacon 12 shown in Figure 1 are also shown in Figure 2. The pre-processing unit 34, frequency discrimination unit 40, spectral feature extraction unit 52, and ground beacon detection unit 60 in Figure 2 constitute the reception processing unit 32 in Figure 1, and the command output unit 72 and output control unit 74 in Figure 2 constitute the control unit 70 in Figure 1. Note that the configuration of the functional block diagram in Figure 2 is an example, and configurations with some of the functional blocks in Figure 2 changed, deleted, or divided, or with other functional blocks added, are also possible. Furthermore, the configuration of the hardware and software that specifically constitute each functional block is not limited to the configuration described below with reference to Figure 2, but may be any configuration within the scope of realizing each function.

First, to explain the basic concept of the ground coil detection method using the ground coil detection device 10, the ground coil detection device 10 extracts a discrete frequency spectrum from the received wave received by the secondary coil 16 of the on-board coil 12 in the receiving processing unit 32, as described later. Furthermore, it performs level conversion based on the received voltage of each frequency component to obtain a spectral waveform SW (SW1, SW2) as shown in Figure 3, which is the envelope of the discrete frequency spectrum. Then, focusing on the waveform shape that has a maximum value (max) at the resonant frequency and a minimum value (min) at the anti-resonant frequency when the on-board coil 12 and the ground coil 80 are electromagnetically coupled, the ground coil 80 and its resonant frequency are detected by extracting a predetermined shape (peak shape PS) that includes that waveform shape.

Returning to Figure 2, the transmission processing unit 22 includes a waveform table 24, a D/A conversion unit 26, and a power amplification unit 28. The waveform table 24 holds transmission wave data of a discrete frequency spectrum in which multiple frequencies are superimposed. These multiple frequencies include all of the multiple predetermined frequencies that the ground beacon 80 to be detected may have, and are set at frequency intervals smaller than the multiple predetermined frequencies. In this embodiment, 64 frequencies with 1 kHz or 2 kHz intervals in the 70 kHz to 139 kHz band are set as multiple frequencies, including f1 to f8 (73 kHz, 80 kHz, 85 kHz, 90 kHz, 95 kHz, 103 kHz, 120 kHz, 130 kHz) shown at the top of Figure 3 as the multiple predetermined frequencies. Such a waveform table 24 is stored in various types of memory, such as flash memory. Furthermore, the bandwidth of the discrete frequency spectrum may be set with a margin particularly on the lower frequency side, so that it includes the anti-resonant frequency of the combined impedance during electromagnetic coupling between the on-board unit 12 and the ground unit 80 for all possible resonant frequencies (multiple predetermined frequencies f1 to f8) that the ground unit 80 may have.

The transmitted wave data read from the waveform table 24 is converted from digital data to analog in the D/A conversion unit 26, and then the power is amplified in the power amplification unit 28 before being input as a transmitted wave to the primary coil 14 of the on-board unit 12. At this time, each frequency component of the discrete frequency spectrum of the transmitted wave is input at a constant voltage level. Figure 4(a) shows a transmitted wave with such a discrete frequency spectrum, input at a constant voltage level, for the low-frequency side. The illustrated frequency band includes f1 and f2 among a plurality of predetermined frequencies f1 to f8. Note that any D/A converter is used in the D/A conversion unit 26, and a known amplifier such as an amplifier is used in the power amplification unit 28.

The magnitude of the frequency interval Δf of the multiple frequencies illustrated in Figure 4(a) is set to an appropriate value considering factors such as accurately detecting the maximum value max (see Figure 3) at the resonant frequency including the frequency error Δe of the ground tonearm 80, and the load on the receiving processing unit 32. For example, Δf is not limited to this, but is set using the relationship "0.5Δe ≤ Δf ≤ Δe" as a guideline. Furthermore, Δf does not need to be the same value within the discrete frequency spectrum; any different value may be set depending on the frequency error Δe of each ground tonearm 80 to be detected, and the interval between multiple predetermined frequencies f1 to f8. Based on these factors, in this embodiment, the frequency interval Δf is set to 1 kHz or 2 kHz as described above.

Returning to Figure 2, the secondary coil 16 of the on-board unit 12 receives a received wave that reflects the electromagnetic interaction with the on-board unit 12, including the effect of loose electromagnetic coupling with the primary coil 14 to which the transmitted wave is input, as described above. This received wave is then taken into the pre-processing unit 34. The pre-processing unit 34 includes an amplifier 36 and an A/D conversion unit 38. The received wave is power-amplified in the amplifier 36 and further converted from an analog waveform to digital data in the A/D conversion unit 38. The received wave data processed in the pre-processing unit 34 is transmitted to the frequency discrimination unit 40. Note that any amplifier, such as a standard amplifier, can be used for the amplifier 36, and any A/D converter can be used for the A/D conversion unit 38.

The frequency discrimination unit 40 includes a quadrature synchronous detection unit 42 and a dB polar coordinate transformation unit 44. The quadrature synchronous detection unit 42 uses synchronous detection (quadrature detection) to extract a discrete frequency spectrum from the received wave data processed by the preceding processing unit 34, identical to the waveform data stored in the waveform table 24 of the transmission processing unit 22. Specifically, in this embodiment, a discrete frequency spectrum with 64 frequency components, where the frequency interval Δf is 1 kHz or 2 kHz, is extracted in the 70 kHz to 139 kHz band. At this time, the average voltage level and phase of each frequency component are obtained. The voltage levels of each frequency component of the discrete frequency spectrum extracted by the quadrature synchronous detection unit 42 are further converted to dB values by the dB polar coordinate transformation unit 44. The frequency discrimination unit 40, which performs such processing, can be implemented, for example, using an FPGA. Note that the frequency discrimination unit 40 may also use a method other than synchronous detection, such as FFT, to extract the discrete frequency spectrum.

Here, Figures 4(b) and (c) show examples of discrete frequency spectra processed by the frequency discrimination unit 40. Figure 4(b) shows the state where coupling with the ground coil 80 does not occur, and Figure 4(c) shows the state where coupling with the ground coil 80 occurs. Note that in Figures 4(b) and (c), some frequency components are omitted from the illustration. Referring to Figure 4(b), it can be seen that the level increases monotonically and linearly as the frequency increases, which is due to the monotonically increasing characteristic corresponding to the uncoupled level envelope UE. The spectral waveform SW, which is defined by the level of each frequency component of the discrete frequency spectrum, has a shape that is mainly based on the impedance frequency characteristics of the coil of the on-board unit 12, along the uncoupled level envelope UE. However, the overall levels of the uncoupled level envelope UE and spectral waveform SW are influenced by a combination of static factors (circuit gain, degree of coupling of the on-board unit 12, and cables to the on-board unit 12) and dynamic factors that occur during the operation of the railway vehicle 86 (such as metal structures between the tracks 84). Therefore, the overall level fluctuations of the uncoupled level envelope UE and spectral waveform SW can reach approximately 45 dB, as indicated by the arrows in Figure 4(b).

On the other hand, referring to Figure 4(c), it can be seen that the spectral waveform SW, defined by the levels of each frequency component of the discrete frequency spectrum, has a shape similar to the spectral waveform SW2 shown in Figure 3, and appears at a level corresponding to the uncoupled level envelope UE. That is, when the on-board unit 12 and the ground unit 80 are electromagnetically coupled, the equivalent circuit becomes a series-parallel circuit of the on-board unit coil and the ground unit coil, and the spectral waveform SW takes on a shape with the characteristics shown in the figure. Specifically, a voltage spectrum is generated with the oscillation frequency of the ground unit 80 as the resonance point (maximum point max), and an anti-resonance point (minimum point min) at a lower frequency. At this time, the stronger the coupling between the on-board unit 12 and the ground unit 80, the greater the level value of the resonance point (maximum point max), but the frequency remains constant. In contrast, the anti-resonance point (minimum point min) becomes smaller and shifts to the lower frequency side as the coupling between the on-board unit 12 and the ground unit 80 becomes stronger. This is illustrated in Figure 3 by the arrow between spectral waveform SW1 and spectral waveform SW2.

Furthermore, the overall level fluctuations of the uncoupled level envelope UE and spectral waveform SW, as explained with reference to Figure 4(b), can also occur in the coupled state with the ground unit 80 shown in Figure 4(c). For example, Figure 3 shows three uncoupled level envelopes UE, and fluctuations in the level of these uncoupled level envelopes UE cause fluctuations in the overall level of the spectral waveform SW. Spectral waveforms SW1 and SW2 in Figure 3 correspond to the middle of the three uncoupled level envelopes UE. Figure 12 also illustrates how level fluctuations of up to approximately 45 dB can occur.

Returning to Figure 2, the discrete frequency spectrum data processed by the frequency discrimination unit 40 is transmitted to the spectral feature extraction unit 52. In this embodiment, the spectral feature extraction unit 52 and the ground signal detection unit 60, which are part of the receiving processing unit 32, and the command output unit 72 and the output control unit 74, which are part of the control unit 70, are implemented by software executed by the CPU 50. Here, the processing time unit of the CPU 50 may be larger than the processing time unit up to the frequency discrimination unit 40, taking into consideration the processing load of the CPU 50. In this case, during one processing time unit of the CPU 50, the data of multiple discrete frequency spectra processed by the frequency discrimination unit 40 will be read at once and processed by the CPU 50.

The spectral feature extraction unit 52 consists of a spectral shape recognition unit 54 and a peak normalization unit 56. The spectral shape recognition unit 54 approximates the spectral waveform SW of the discrete frequency spectrum obtained from the frequency discrimination unit 40 into a piecewise linear shape using three types of line segments: positive gradient line segments, negative gradient line segments, and flat line segments. For example, Figure 5 shows how a spectral waveform SW similar in shape to the spectral waveform SW in Figure 4(c) is approximated into a piecewise linear spectral waveform SW'. That is, a positive gradient line segment P is a line segment whose level increases toward the high-frequency side (right side in the figure), a negative gradient line segment M is a line segment whose level decreases toward the high-frequency side, and a flat line segment F is a relatively nearly horizontal line segment that does not correspond to a positive gradient line segment P or a negative gradient line segment M.

In Figure 5, the spectral waveform SW is approximated by seven line segments between points D1 and D8. Compared to the spectral waveform SW defined from a discrete frequency spectrum with 64 frequency components, the amount of data is compressed. The type and number of line segments used in approximating the spectral waveform SW are determined considering factors such as the shape of the spectral waveform SW and the monotonically increasing characteristics of the uncoupled level envelope UE. Furthermore, while the vertical and horizontal axes are omitted in Figure 5, and in Figures 6, the upper part of Figure 9, Figures 10(b) and (c), 11, and 12 (described later), in all of these, the vertical axis represents level (higher towards the top) and the horizontal axis represents frequency (higher towards the right).

Furthermore, the spectral shape recognition unit 54 calculates a primary attribute group for each of the line segments (seven line segments in the example of Figure 5) that constitute the piecewise spectral waveform SW'. The primary attribute group for each line segment in this embodiment includes the frequency, level, and phase of the starting point (the endpoint on the low-frequency side) of each line segment, and the frequency, level, and phase of the ending point (the endpoint on the high-frequency side) of each line segment. The frequency and level are the frequency and level at which the starting or ending point of each line segment is located, and the phase is the phase in the frequency component of the starting or ending point of each line segment obtained when the discrete frequency spectrum is extracted by the frequency discrimination unit 40. For example, the positive gradient line segment P, located third from the left in Figure 5, has its starting point D3 and ending point D4, and its frequency, level, and phase at the starting point D3 and its ending point D4 are calculated as described above.

Furthermore, the spectral shape recognition unit 54 calculates a secondary line segment attribute group for each of the line segments constituting the piecewise spectral waveform SW' using the primary line segment attribute group. The secondary line segment attribute group in this embodiment includes level difference, bandwidth, and phase difference. Level difference is the difference between the level at the starting point and the level at the ending point of each line segment. Bandwidth is the difference between the frequency at the starting point and the frequency at the ending point of each line segment, and phase difference is the difference between the phase at the starting point and the phase at the ending point of each line segment. For example, for the negative gradient line segment M located second from the right in Figure 5, the level difference is calculated by subtracting the level at the starting point D6 from the level at the ending point D7; the bandwidth is calculated by subtracting the frequency at the starting point D6 from the frequency at the ending point D7; and the phase difference is calculated by subtracting the phase at the starting point D6 from the phase at the ending point D7. Each attribute calculated by the spectral shape recognition unit 54 is retained for subsequent processing.

Next, the peak normalization unit 56 reads the spectral waveform SW' approximated as a piecewise linear shape by the spectral shape recognition unit 54 and extracts the peak shape PS as shown in Figure 6. That is, the peak shape PS is the portion where the waveform changes from a positive gradient line segment P to a negative gradient line segment M towards the higher frequency side (right side in the figure), as shown in Figure 6(a). Furthermore, the peak shape PS may also be the portion where the waveform changes from a positive gradient line segment P to a flat line segment F and then to a negative gradient line segment M towards the higher frequency side (right side in the figure), as shown in Figure 6(b), and where the bandwidth (length in the left-right direction in the figure) of the flat line segment F is a relatively small bandwidth. Note that the peak shape PS can be considered a candidate for the predetermined shape mentioned with reference to Figure 3.

Furthermore, prior to extracting the peak shape PS as described above, if the piecewise approximated spectral waveform SW' contains a portion where a relatively small bandwidth (very small bandwidth) flat line segment F is present between two positive gradient line segments P, these two positive gradient line segments P and flat line segment F may be integrated and treated as a single positive gradient line segment P. Similarly, if a portion contains a relatively small bandwidth (very small bandwidth) flat line segment F between two negative gradient line segments M, these two negative gradient line segments M and flat line segment F may be integrated and treated as a single negative gradient line segment M.

For example, in the example in Figure 5, assuming that the bandwidth of the fourth flat line segment F from the left is relatively small, points D3 to D6 are integrated as a single positive gradient line segment P, and points D3 to D7 are extracted as the peak shape PS. The determination of which flat line segments F should have a small bandwidth to be included in the peak shape PS as shown in Figure 6(b), or which flat line segments F should have a small bandwidth to be integrated with two positive gradient line segments P or two negative gradient line segments M, should be performed using parameters based on experimental results. Note that multiple peak shapes PS may be extracted simultaneously from the piecewise spectral waveform SW', and if the conditions are not met, no peak shape PS may be extracted. The number of peak shapes PS extracted from the piecewise spectral waveform SW' is retained for subsequent processing.

Furthermore, the peak normalization unit 56 calculates a group of normalized peak primary attributes for each of the peak shapes PS extracted from the piecewise spectral waveform SW'. The group of normalized peak primary attributes in this embodiment includes the normalized peak level, normalized peak bandwidth, normalized peak frequency, normalized peak absolute level, and normalized peak phase difference. As shown in Figure 6, the normalized peak level NL is the sum of the absolute value of the level difference of the positive gradient line segment P that constitutes each peak shape PS and the absolute value of the level difference of the negative gradient line segment M that constitutes each peak shape PS. In Figure 6(a), for ease of understanding, the sign of the negative gradient line segment M from point D11 to point D12 is reversed (because it is an absolute value) and shown as a dashed line as the negative gradient line segment M' from point D11 to point D12', and the level from point D10 to point D12' corresponds to the normalized peak level NL. Similarly, in Figure 6(b), the negative gradient line segment M from point D17 to point D18 is reversed in sign and shown as a dashed line as the negative gradient line segment M' from point D17 to point D18'. The level from point D15 to point D18' corresponds to the normalized peak level NL.

The normalized peak bandwidth NB is the sum of the bandwidths of the positive gradient line segments P and the negative gradient line segments M that constitute each peak shape PS, as shown in Figure 6(a). Furthermore, if the peak shape PS includes a flat line segment F with a small bandwidth, the normalized peak bandwidth NB is the sum of the bandwidths of the positive gradient line segment P, the flat line segment F, and the negative gradient line segment M, as shown in Figure 6(b). The normalized peak frequency NF is the average of the endpoint frequency of the positive gradient line segment P and the starting frequency of the negative gradient line segment M that constitutes each peak shape PS; for example, it is the sum of these two frequencies divided by 2.

The normalized peak absolute level (NAL) is the average of the levels at the endpoints of the positive gradient line segments P that constitute each peak shape PS and the levels at the starting points of the negative gradient line segments M that constitute each peak shape PS. For example, it is the sum of these two levels divided by 2. The normalized peak phase difference is the sum of the phase differences of the positive gradient line segments P that constitute each peak shape PS and the phase differences of the negative gradient line segments M that constitute each peak shape PS. Furthermore, if the peak shape PS includes a flat line segment F with a small bandwidth, as shown in Figure 6(b), the normalized peak phase difference is the sum of the phase differences of the positive gradient line segment P, the flat line segment F, and the negative gradient line segment M.

For example, in the example in Figure 5, if points D3 to D7 are extracted as the peak shape PS, then each attribute of the normalized peak primary attribute group is calculated for this peak shape PS as described above. Note that the level difference of each line segment used in calculating the normalized peak level NL, the bandwidth of each line segment used in calculating the normalized peak bandwidth NB, and the phase difference of each line segment used in calculating the normalized peak phase difference are included in the line segment secondary attribute group calculated by the spectral shape recognition unit 54. Furthermore, the frequency of the endpoint of the positive gradient line segment P and the frequency of the starting point of the negative gradient line segment M, used in calculating the normalized peak frequency NF, and the level of the endpoint of the positive gradient line segment P and the level of the starting point of the negative gradient line segment M, used in calculating the normalized peak absolute level NAL, are included in the line segment primary attribute group calculated by the spectral shape recognition unit 54.

Furthermore, the peak normalization unit 56 calculates a normalized peak secondary attribute group for each of the peak shapes PS extracted from the piecewise spectral waveform SW'. The normalized peak secondary attribute group in this embodiment includes the normalization gradient and the normalization gradient level distribution ratio. The normalization gradient NS is the ratio of the normalized peak level NL and the normalized peak bandwidth NB included in the normalized peak primary attribute group, and is calculated, for example, by dividing the normalized peak level NL by the normalized peak bandwidth NB (see the bottom of Figure 9). The normalization gradient level distribution ratio is the ratio of the absolute value PL of the level difference of the positive gradient line segment P constituting each peak shape PS (see Figures 10(b) and (c)) to the absolute value ML of the level difference of the negative gradient line segment M constituting each peak shape PS (see Figures 10(b) and (c)), and the level difference of each line segment is included in the line segment secondary attribute group calculated by the spectral shape recognition unit 54. For example, the normalized gradient level distribution ratio is calculated by dividing the absolute value of the level difference of a positive gradient line segment P by the sum of the absolute values of the level differences of the positive gradient line segment P and the negative gradient line segment M.

Referring to Figure 2, the normalized peak primary attribute group and normalized peak secondary attribute group calculated by the spectral feature extraction unit 52 as described above are transmitted to the ground beacon detection unit 60. The ground beacon detection unit 60 detects that the on-board unit 12 has electromagnetically coupled with the ground beacon 80, and the resonant frequency of the electromagnetically coupled ground beacon 80. It includes a ground beacon candidate determination unit 62 and a ground beacon final determination unit 64. The ground beacon candidate determination unit 62 selects candidate peak shapes PS from the peak shapes PS extracted by the spectral feature extraction unit 52 that indicate coupling with the ground beacon 80 has occurred. The ground beacon final determination unit 64 performs a final determination of whether the peak shape PS selected as a candidate indicating coupling with the ground beacon 80 truly indicates coupling with the ground beacon 80, and estimates whether equipment or environmental abnormalities have occurred.

Figure 7 shows a flowchart illustrating the specific operation of the ground beacon candidate determination unit 62, and Figure 8 shows a flowchart illustrating the specific operation of the ground beacon final determination unit 64. The following explanation will follow these flowcharts. Note that the flowcharts shown in Figures 7 and 8 are examples of specific operations and do not limit the operation of the ground beacon candidate determination unit 62 and the ground beacon final determination unit 64. Therefore, depending on the configuration and circumstances of the ground beacon detection device 10, for example, some of the steps shown in Figures 7 and 8 may be deleted, modified, or added as appropriate.

First, with reference to Figure 7, the operation of the ground tonearm candidate determination unit 62 will be explained.
S10 (Data acquisition): The number of peak shapes PS extracted by the peak normalization unit 56, and the normalized peak primary attribute group and normalized peak secondary attribute group for each peak shape PS are acquired from the peak normalization unit 56.
S20 (Counter Reset): Resets the ground beacon candidate counter, which counts the number of candidates for peak shape PS indicating that coupling with ground beacon 80 has occurred, and the processing counter, which counts the number of times the ground beacon determination process described later has been repeated. Here, the ground beacon candidate counter is WCC and the processing counter is i, and zero is assigned to the ground beacon candidate counter WCC and 1 is assigned to the processing counter i.

S30 (Processing Counter Determination): The process determines whether the processing counter i is less than or equal to the number of peak shape PSs. If the process determines that the processing counter i is less than or equal to the number of peak shape PSs (YES), then there are still peak shape PSs for which the ground coil determination process described later has not been performed, and the process proceeds to S40. If the process determines that the processing counter i is not less than or equal to the number of peak shape PSs (NO), then the process by the ground coil candidate determination unit 62 ends because the ground coil determination process described later has been completed for all extracted peak shape PSs, or no peak shape PSs have been extracted.

S40 (Normalized Peak Level Determination): The normalized peak level NL of the i-th peak shape PS among the extracted peak shapes PS is determined to be within a predetermined range. This predetermined range is the appropriate range that the normalized peak level NL of the peak shape PS generated by electromagnetic coupling with the ground beacon 80 can take, and is determined based on experimental results, etc. If it is determined that the normalized peak level NL is within the predetermined range (YES), the process proceeds to S50 to continue the ground beacon candidate determination. If it is determined that the normalized peak level NL is not within the predetermined range (NO), the process proceeds to S110.

S50 (Normalized Peak Bandwidth Determination): The normalized peak bandwidth NB of the i-th peak shape PS among the extracted peak shapes PS is determined to be within a predetermined range. This predetermined range is the appropriate range that the normalized peak bandwidth NB of the peak shape PS generated by electromagnetic coupling with the ground beacon 80 can take, and is determined based on experimental results, etc. If it is determined that the normalized peak bandwidth NB is within the predetermined range (YES), the process proceeds to S60 to continue the ground beacon candidate determination. If it is determined that the normalized peak bandwidth NB is not within the predetermined range (NO), the process proceeds to S110.

S60 (Normalization Gradient Determination): This step determines whether the normalization gradient NS of the i-th peak shape PS among the extracted peak shapes PS falls within a predetermined range. This predetermined range is the appropriate range that the normalization gradient NS of the peak shape PS generated by electromagnetic coupling with the ground beacon 80 can take, and is determined based on experimental results, etc. If it is determined that the normalization gradient NS is within the predetermined range (YES), the process proceeds to S70 to continue the ground beacon candidate determination. If it is determined that the normalization gradient NS is not within the predetermined range (NO), the process proceeds to S110.

Referring to Figure 9, the waveform at the top of Figure 9 shows the peak shape PS caused by noise on the left and the peak shape PS caused by ground beacon coupling on the right. From this, it can be seen that the normalized absolute peak level NAL of the peak shape PS caused by noise is greater than the normalized absolute peak level NAL of the peak shape PS caused by ground beacon coupling. Thus, it is conceivable that the absolute value of the level of the peak shape PS caused by noise may be greater than that of the peak shape PS caused by ground beacon coupling. Therefore, as shown at the bottom of Figure 9, we compare the normalized peak level NL, normalized peak bandwidth NB, and normalized gradient NS of both peak shapes PS. Here, the normalized gradient NS is obtained by dividing the normalized peak level NL by the normalized peak bandwidth NB. As a result, the normalized peak level NL and normalized peak bandwidth NB are greater for the ground beacon coupling than for the noise. Furthermore, the normalized gradient NS is greater for noise than for ground beacon coupling, and the slope is gentler for ground beacon coupling than for noise. In steps S40, S50, and S60 above, this relationship is used to differentiate between noise and ground-level signal coupling.

S70 (Normalized Gradient Level Distribution Ratio Determination): This step determines whether the normalized gradient level distribution ratio of the i-th peak shape PS among the extracted peak shapes PS falls within a predetermined range. This predetermined range is the appropriate range that the normalized gradient level distribution ratio of the peak shapes PS generated by electromagnetic coupling with the ground beacon 80 can take, and is determined based on experimental results, etc. If it is determined that the normalized gradient level distribution ratio is within the predetermined range (YES), the process proceeds to S80 to continue the ground beacon candidate determination. If it is determined that the normalized gradient level distribution ratio is not within the predetermined range (NO), the process proceeds to S110.

Referring to Figure 10(a), as the on-board unit 12 approaches the ground unit 80 due to the movement of the railway vehicle 86, a main lobe 90 and side lobes 92 are generated due to the influence of magnetic flux corresponding to their relative positions. The main lobe 90 is to be detected in order to detect coupling with the ground unit 80. As shown in Figure 10(b), the peak shape PS generated at this time has its anti-resonant frequency (minimum) located lower than the resonant frequency (maximum). Therefore, as shown in the figure, the absolute value PL of the level difference of the positive gradient line segment P is greater than the absolute value ML of the level difference of the negative gradient line segment M. Consequently, the normalized gradient level distribution ratio, which is the ratio of these two values, is greater than 0.5 when calculated using "PL / (PL + ML)".

On the other hand, the side lobe 92 occurs twice, before and after the main lobe 90, but it should not be detected as a coupling with the ground sensor 80. As shown in Figure 10(c), the peak shape PS generated by the side lobe 92 has its minimum point (min) at a higher frequency than the maximum point (max) at the resonant frequency. Due to the influence of the magnetic flux direction, the relative magnitudes of the resonant and anti-resonant frequencies are reversed compared to the main lobe 90. Therefore, as shown in the figure, the absolute value ML of the level difference of the negative gradient line segment M is greater than the absolute value PL of the level difference of the positive gradient line segment P. Consequently, the normalized gradient level distribution ratio, which is the ratio of these two values, is less than or equal to 0.5 when calculated using "PL / (PL + ML)". In the above-described S70, this characteristic of the normalized gradient level distribution ratio is used to separate the main lobe 90 from the side lobe 92.

S80 (Normalized Peak Frequency Determination): This step determines whether the normalized peak frequency NF of the i-th peak shape PS among the extracted peak shapes PS falls within a predetermined range. This predetermined range is the appropriate range that the normalized peak frequency NF of the peak shape PS generated by electromagnetic coupling with the ground beacon 80 can take, and is determined based on experimental results, etc. If it is determined that the normalized peak frequency NF is within the predetermined range (YES), the process proceeds to S90, as all the ground beacon determination processes in the flowchart of Figure 7 have been satisfied. If it is determined that the normalized peak frequency NF is not within the predetermined range (NO), the process proceeds to S110.

S90 (Ground coil candidate determination): Of the extracted peak shapes PS, the i-th peak shape PS is determined to be a candidate indicating a connection with ground coil 80, and a flag is set to the i-th peak shape PS.
S100 (Ground coil candidate counter increment): Increment the ground coil candidate counter WCC by 1.
S110 (Rejection of Ground Coil Candidate): It is determined that the i-th peak shape PS among the extracted peak shapes PS is not a candidate indicating a connection with Ground Coil 80.
S120 (Processing counter increment): The processing counter i is incremented by 1. Then, the process returns to S30 to determine whether to continue the ground signal determination process. In other words, steps S40 to S120 are repeatedly executed until the determination process for all detected peak shapes PS is completed.

Next, with reference to Figure 8, the operation of the ground tone final determination unit 64 will be explained.
S200 (Peak Shape Count Determination): This step determines whether the number of peak shapes PS obtained in S10 of Figure 7 is equal to or greater than a predetermined number. The predetermined number is an appropriate value for distinguishing between noise that generates multiple peak shapes PS simultaneously and ground-level coupling, and is determined based on experimental results, etc. If it is determined that the number of peak shapes PS is equal to or greater than the predetermined number (YES), the process proceeds to S210. If it is determined that the number of peak shapes PS is less than the predetermined number (NO), the process proceeds to S220.

Figure 11 illustrates several peak shapes PS that are thought to be caused by noise. Pulse noise originating from the power supply system, in particular, tends to occur over a wide bandwidth, albeit for a short duration, resulting in the simultaneous extraction of multiple peak shapes PS as shown in the figure. Among these, it is possible to obtain a peak shape PS similar in shape to that observed during ground coil coupling, such as the second peak shape PS from the left. However, if multiple peak shapes PS are extracted at other frequencies simultaneously, there is a high probability that all simultaneously extracted peak shapes PS are due to noise, and they can be considered candidates for ground coil exclusion. In S200 described above, this is used to differentiate between noise and ground coil coupling.

S210 (No ground coils and abnormal environmental conditions): In S200, it was determined that more than the number of peak shapes PS that could be attributed to noise were extracted simultaneously. Therefore, it is determined that none of the extracted peak shapes PS indicate ground coil coupling. Furthermore, since noise is present, it is determined that there is an abnormal environmental condition.
S220 (First determination of ground coil candidate counter): Determine whether the ground coil candidate counter WCC is 3 or greater. If the ground coil candidate counter WCC is 3 or greater (YES), proceed to S230; if the ground coil candidate counter WCC is 2 or less (NO), proceed to S260.

S230 (Ground Tone Candidate Counter Second Determination): This step determines whether the ground tonenea candidate counter WCC has remained at 3 or higher for an extended period. "Extended period" here refers to a situation where, despite the receiving processing unit 32 repeatedly reading new received waves from the secondary coil 16 of the on-board tonenea 12 and extracting discrete frequency spectra and peak shape PS, the ground tonenea candidate counter WCC remains at 3 or higher throughout this process. The time (number of processing cycles) used for this determination is set based on experiments, etc. If it is determined that the above state has continued for an extended period (YES), the process proceeds to S240; if it is determined that it has not continued for an extended period (NO), the process proceeds to S250. Note that the value "3" used in the determination of the ground tonenea candidate counter WCC in S220 and this step S230 is just an example, and other values may be used.

S240 (No ground beacon and equipment malfunction): Since the ground beacon candidate counter WCC has remained at 3 or higher for an extended period, it is presumed that this is due to a malfunction or other issue in one of the parts of the ground beacon detection device 10, and therefore an equipment malfunction is detected. Furthermore, the count result of the ground beacon candidate counter WCC is determined to be due to an equipment malfunction and not to the peak shape PS indicating ground beacon coupling, and therefore no ground beacon coupling is detected.
S250 (No ground beacon and environmental anomaly present): Although the ground beacon candidate counter WCC has not remained at 3 or higher for an extended period, the presence of three or more peak shapes PS, which are considered candidates for ground beacon coupling, is judged to be an anomaly. Therefore, it is determined that environmental anomalies such as noise are occurring, and that ground beacon coupling has not been detected.

S260 (Ground coil candidate counter third determination): Determine whether the ground coil candidate counter WCC is 2 or not. If the ground coil candidate counter WCC is 2 (YES), proceed to S270; if the ground coil candidate counter WCC is 1 or less (NO), proceed to S280.
S270 (Ground coil detection): Of the two peak shapes PS that were flagged as candidates indicating ground coil coupling in S90 of Figure 7, the peak shape PS with the larger normalized peak level NL is determined to indicate coupling with ground coil 80.

S280 (Ground coil candidate counter 4th determination): Determine whether the ground coil candidate counter WCC is 1 or not. If the ground coil candidate counter WCC is 1 (YES), proceed to S290; if the ground coil candidate counter WCC is not 1, i.e., zero (NO), proceed to S300.
S290 (Ground coil detection): In S90 of Figure 7, one peak shape PS that has been flagged as a candidate indicating ground coil coupling is determined to indicate coupling with ground coil 80.

S300 (No ground beacon): In S90 of Figure 7, there is no peak shape PS that has been flagged as a candidate indicating ground beacon coupling, so it is determined that ground beacon 80 was not detected.
S310 (Information Transmission): The information determined in the step immediately preceding this step is transmitted to the control unit 70. Specifically, it is transmitted whether or not the ground beacon 80 was detected, if detected, the predetermined frequency of that ground beacon 80, whether or not there is an environmental abnormality, whether or not there is an equipment abnormality, and the normalized peak primary attribute group and normalized peak secondary attribute group for each peak shape PS (especially peak shapes PS that are attributed to noise or equipment abnormalities). With this processing complete, the processing by the ground beacon final determination unit 64 is finished.

Returning to Figure 2, the command output unit 72 of the control unit 70 outputs train control information to the output control unit 74 based on information acquired from the ground beacon detection unit 60 and speed checks. Specifically, when the ground beacon detection unit 60 notifies that a ground beacon 80 has been detected, it acquires the resonant frequency of the detected ground beacon 80, identifies the information indicated by that resonant frequency (such as a stop indication or speed indication), and reflects it in the train control information. Furthermore, the output control unit 74 outputs control signals to the I/F circuit 76 for specific brake control and speed control by the vehicle control unit 78 (see Figure 1) based on the train control information from the command output unit 72.

Furthermore, the control unit 70 may perform arbitrary processing for the purpose of assisting existing functions or enhancing functionality, using various attributes calculated by the spectral feature extraction unit 52, namely, attributes included in the line segment primary attribute group, line segment secondary attribute group, normalized peak primary attribute group, and normalized peak secondary attribute group, as well as the number of extracted peak shapes PS and the number of peak shapes PS counted as ground tone candidates. For example, the control unit 70 may perform noise analysis using the normalized peak primary attribute group and normalized peak secondary attribute group of the peak shapes PS determined to be due to noise in S210 and S250 of Figure 8. Similarly, the control unit 70 may perform equipment malfunction analysis using various attributes when equipment malfunction is determined to have occurred in S240 of Figure 8.

Herein, the ground coil detection device 10 according to the embodiment of the present invention is not limited to the configuration shown in Figures 1 to 12, but may have a different configuration. For example, the ground coil detection device 10 is not limited to the ground coil 80 of the ATS, but may detect ground coils 80 of other railway equipment. Also, the multiple predetermined frequencies that the ground coil 80 may have as resonant frequencies are not limited to f1 to f8 in Figure 3, but may be other frequencies, more frequencies, or fewer frequencies. Accordingly, the discrete frequency spectrum of the transmitted wave generated by the transmission processing unit 22 is not limited to a bandwidth of 70 kHz to 139 kHz, a frequency interval of 1 kHz or 2 kHz, and 64 frequencies, but may be a different bandwidth, a different frequency interval, and a different number of frequencies. Furthermore, the configurations of the line segment primary attribute group, line segment secondary attribute group, normalized peak primary attribute group, and normalized peak secondary attribute group are not limited to those described above, and may be configurations that do not include any of the above-described attributes, or configurations that include other attributes. Also, the calculation method for each attribute may differ from that described above.

Now, according to the embodiment of the present invention having the above configuration, the following effects can be obtained. That is, as shown in Figures 1 and 2, the ground beacon detection device 10 according to the embodiment of the present invention detects a ground beacon 80 from a moving railway vehicle 86, and includes an on-board unit 12, a transmission processing unit 22, a pre-processing unit 34, a frequency discrimination unit 40, a spectral feature extraction unit 52, and a ground beacon detection unit 60. The on-board unit 12 has a primary coil 14 and a secondary coil 16 that are loosely coupled to each other and constantly transmit and receive signals between them. When the railway vehicle 86 is moving, if the on-board unit 12 attached to the railway vehicle 86 approaches a ground beacon 80 near the track 84, the ground beacon 80 and the on-board unit 12 are electromagnetically coupled. The transmission processing unit 22 injects a transmission wave having a discrete frequency spectrum into the primary coil 14 of the on-board unit 12. The discrete frequency spectrum of the transmitted wave includes multiple frequencies that encompass all of the multiple predetermined frequencies (e.g., f1 to f8 in Figure 3) that the detected ground tonearm 80 may have as its resonant frequency, with a frequency interval Δf smaller than the multiple predetermined frequencies (see Figure 4(a)).

The transmission wave, as described above, injected into the primary coil 14 is received by the secondary coil 16, which is loosely coupled to the primary coil 14. The frequency discrimination unit 40 extracts a discrete frequency spectrum from the received wave, which has the same frequency components as the one generated by the transmission processing unit 22, after it has been received by the secondary coil 16 and pre-processed by the pre-processing unit 34. The spectral feature extraction unit 52 then generates a spectral waveform SW (see Figures 3 and 4) from the discrete frequency spectrum extracted by the frequency discrimination unit 40, defined by the levels of each of those frequency components, and further extracts a predetermined shape from the spectral waveform SW. The ground coil detection unit 60 then detects the ground coil 80 depending on the presence or absence of the predetermined shape extracted from the spectral waveform SW.

Here, when the ground coil 80 and the on-board coil 12 are electromagnetically coupled as described above, one of the multiple predetermined frequencies that each on-board coil 80 possesses as information is transmitted to the on-board coil 12 as the resonant frequency. That is, the resonant frequency affects the received wave received by the secondary coil 16 of the on-board coil 12, and a shape appears in the spectral waveform SW generated based on the received wave in which the level of the frequency component with the same frequency as the resonant frequency of the ground coil 80 becomes large. Therefore, by setting a predetermined shape that the level becomes large at the resonant frequency as the shape extracted from the spectral waveform SW by the spectral feature extraction unit 52, it is possible to detect that the on-board coil 12 has been electromagnetically coupled to the ground coil 80 having that resonant frequency. As a result, the resonant frequency appearing in the spectral waveform SW can be extracted by the shape of the waveform rather than the magnitude of the level, so it is possible to suppress false detection of simply large noise levels and the effects of absolute level fluctuations due to various factors (see Figure 12), thereby improving the reliability of ground coil detection. Furthermore, extracting a predetermined shape from the spectral waveform SW can be achieved using a relatively simple algorithm by simplifying the processing data and employing other techniques.

Furthermore, in the embodiment of the present invention, the ground coil detection device 10 injects the transmitted wave into the primary coil 14 of the on-board coil 12 at a constant voltage when the transmission processing unit 22 injects the transmitted wave. The frequency discrimination unit 40 then performs level conversion on the discrete frequency spectrum extracted from the received wave of the secondary coil 16 based on the received voltage of each frequency component. Therefore, the spectral feature extraction unit 52 extracts a predetermined shape from the spectral waveform generated based on the discrete frequency spectrum level-converted by the frequency discrimination unit 40. As a result, voltage differences between frequency components that were not present in the transmitted wave appear in the received wave due to electromagnetic coupling with the ground coil 80, and these can be further amplified by a predetermined level conversion operation. Therefore, it becomes easier to extract a predetermined shape from the spectral waveform in which such voltage differences are reflected in the level, thereby improving the detection accuracy of the ground coil 80.

Furthermore, in the embodiment of the present invention, the ground coil detection device 10 has a bandwidth that includes the anti-resonant frequency of the combined impedance during electromagnetic coupling between the on-board coil 12 and the ground coil 80, in the discrete frequency spectrum of the transmitted wave generated by the transmission processing unit 22 and the discrete frequency spectrum extracted by the frequency discrimination unit 40. Since the anti-resonant frequency appears as a shape where the level is lower than the resonant frequency during electromagnetic coupling between the on-board coil 12 and the ground coil 80, the bandwidth of the discrete frequency spectrum is set so that the above-mentioned region below the resonant frequency (predetermined frequency) is covered for all of the multiple predetermined frequencies that the ground coil 80 may have. In the example in Figure 3, the bandwidth is set to include the anti-resonant frequencies corresponding to the resonant frequencies from f1 (73 kHz) to f8 (130 kHz). By including the shape showing the effect of the anti-resonant frequency during electromagnetic coupling in the predetermined shape extracted by the spectral feature extraction unit 52, the distinction between the ground coil 80 and noise becomes clearer, thus further improving the reliability of ground coil detection.

Furthermore, the ground coil detection device 10 according to the embodiment of the present invention more specifically identifies the predetermined shape extracted from the spectral waveform SW by the spectral feature extraction unit 52 as follows. That is, the predetermined shape is set to a shape like the peak shape PS shown in Figure 3, where the level rises from a relatively low anti-resonant frequency towards the resonant frequency, and then decreases towards the resonant frequency. This allows for the setting of a characteristic shape that appears during electromagnetic coupling with the ground coil 80, where the level rises significantly from the anti-resonant frequency to the resonant frequency before decreasing. This makes it even easier to distinguish between noise and the ground coil, and enables more accurate detection of the ground coil 80.

Furthermore, in the embodiment of the present invention, the ground signal detection device 10, as shown in Figure 5, has a spectral feature extraction unit 52 that approximates the spectral waveform SW, defined by the levels of the frequency components of the discrete frequency spectrum, into a piecewise spectral waveform SW' by combining arbitrary line segments from three types of line segments. These three types of line segments are a positive gradient line segment P whose level increases towards higher frequencies, a negative gradient line segment M whose level decreases towards higher frequencies, and a flat line segment F that does not fall under either the positive gradient line segment P or the negative gradient line segment M. This allows for a reduction in the amount of data processed in subsequent stages compared to directly handling the spectral waveform SW, thus simplifying the implementation of programs for subsequent processing.

Furthermore, in the ground signal detection device 10 according to the embodiment of the present invention, the spectral feature extraction unit 52 calculates a primary attribute group and a secondary attribute group for each line segment of the polylinearly approximated spectral waveform SW'. The primary attribute group includes at least the frequency and level of the starting point, which is the low-frequency endpoint of each line segment, and the frequency and level of the ending point, which is the high-frequency endpoint of each line segment. The secondary attribute group includes at least the level difference between the starting point and the ending point of each line segment, and the bandwidth, which is the frequency difference between the starting point and the ending point of each line segment. This allows the characteristics of each line segment to be represented as a more easily processed data set.

Furthermore, in the embodiment of the present invention, the ground signal detection device 10 has a spectral feature extraction unit 52 that extracts a peak shape PS as shown in Figure 6 from the piecewise approximated spectral waveform SW'. This peak shape PS is the portion that changes from a positive gradient line segment P to a negative gradient line segment M toward the higher frequency side, as shown in Figure 6(a), and/or the portion that changes from a positive gradient line segment P to a negative gradient line segment M via a flat line segment F with a small bandwidth toward the higher frequency side, as shown in Figure 6(b). The flat line segment F with a small bandwidth here is a flat line segment F with a relatively small bandwidth, and whether or not it is a flat line segment F with a small bandwidth is determined by considering the bandwidth of the positive gradient line segment P and the negative gradient line segment M before and after it, as well as experimental results. By processing using this peak shape PS, the amount of data processed in subsequent stages can be compressed compared to handling the entire spectral waveform SW' which is approximated as a piecewise linear waveform. This simplifies the algorithms used in processing by the spectral feature extraction unit 52 and the ground signal detection unit 60, thereby reducing the load on the CPU 50.

Furthermore, in the ground signal detection device 10 according to the embodiment of the present invention, the spectral feature extraction unit 52 calculates the following normalized peak primary attribute group for each extracted peak shape PS, when the positive gradient line segment P constituting the peak shape PS is defined as the peak positive gradient line segment P, and the negative gradient line segment M constituting the peak shape PS is defined as the peak negative gradient line segment M. That is, the normalized peak primary attribute group includes at least the normalized peak level NL, the normalized peak bandwidth NB, and the normalized peak frequency NF, as shown in Figure 6. The normalized peak level NL is the sum of the absolute value of the level difference between the peak positive gradient line segment P and the absolute value of the level difference between the peak negative gradient line segment M. The normalized peak bandwidth NB is the sum of the bandwidth of the peak positive gradient line segment P and the bandwidth of the peak negative gradient line segment M, and if the peak shape PS includes a flat line segment F with a small bandwidth, it is also taken into account in the calculation. Furthermore, the normalized peak frequency NF is the average of the frequency at the end point of the peak positive gradient line segment P and the frequency at the start point of the peak negative gradient line segment M.

The ground coil detection unit 60 then uses the normalized primary peak attribute group of each peak shape PS described above to determine whether the peak shape PS represents the ground coil 80, that is, whether the peak shape PS appears due to electromagnetic coupling between the ground coil 80 and the on-board coil 12 (see S40, S50, and S80 in Figure 7). By using the normalized primary peak attribute group in this way, characteristic peak shapes PS, including the resonant and anti-resonant frequencies that change moment by moment depending on the distance between the on-board coil 12 and the ground coil 80, can be easily extracted using software. Furthermore, since all the attributes included in the normalized primary peak attribute group are simple indicators, the software processing for determining the peak shape PS representing the ground coil 80 is simple and can be processed by a low-speed CPU 50.

Furthermore, since the normalized peak level NL, one of the normalized peak primary attribute groups, is calculated by integrating the positive gradient line segment P and the negative gradient line segment M, a larger signal-to-noise ratio can be obtained compared to conventional peak levels that only use the maximum value detected from simple spectral analysis such as FFT. This enables a highly reliable detection method for the ground transponder 80 with excellent noise immunity, and also allows for lower power consumption in the power amplification unit 28 used in the transmission processing unit 22 compared to conventional methods, contributing to miniaturization and cost reduction of the device. Moreover, because the normalized peak level NL is a relative level that integrates the level difference of the positive gradient line segment P and the level difference of the negative gradient line segment M, the absolute levels cancel each other out, and the effects of circuit gain, on-board transponder coupling degree, and other factors that cause fluctuations in the absolute level do not appear (see Figure 12). Therefore, the reliability of ground transponder detection can be further improved.

Furthermore, in the embodiment of the present invention, the ground coil detection device 10 has a spectral feature extraction unit 52 that calculates a normalized peak secondary attribute group for each extracted peak shape PS. This normalized peak secondary attribute group includes a normalization gradient NS (see Figure 9), which is the ratio of the normalized peak level NL to the normalized peak bandwidth NB. The ground coil detection unit 60 then uses the normalization gradient NS of each peak shape PS calculated by the spectral feature extraction unit 52 to determine whether or not that peak shape PS represents a ground coil 80 (see S60 in Figure 7).

In other words, as shown in Figure 9, the normalization gradient NS, which is the ratio of the normalized peak level NL to the normalized peak bandwidth NB, tends to be smaller (gentler slope) for ground benton coupling than for single-frequency noise when calculated by dividing the normalized peak level NL by the normalized peak bandwidth NB. Therefore, even if the noise frequency and the ground benton frequency are the same, the above-mentioned relationship in the magnitude of the normalization gradient NS allows for accurate detection of the ground benton 80 without misidentifying single-frequency noise as the ground benton 80.

Furthermore, in the embodiment of the present invention, the ground coil detection device 10 has a spectral feature extraction unit 52 that calculates a normalized gradient level distribution ratio for each peak shape PS as a normalized peak secondary attribute group. This normalized gradient level distribution ratio is the ratio of the absolute value PL of the level difference of the peak positive gradient line segment P to the absolute value ML of the level difference of the peak negative gradient line segment M, as shown in Figures 10(b) and (c). The ground coil detection unit 60 then uses the normalized gradient level distribution ratio of each peak shape PS calculated by the spectral feature extraction unit 52 to determine whether or not that peak shape PS represents a ground coil 80 (see S70 in Figure 7).

Here, when the on-board unit 12 and the ground unit 80 are electromagnetically coupled, as shown in Figure 10(a), due to the influence of magnetic flux, side lobes 92 that should not be detected are generated before and after the main lobe 90 that should be detected. As shown in Figure 10(b), in the main lobe 90, a peak shape PS is generated in which the anti-resonant frequency is smaller than the resonant frequency, whereas in the side lobe 92, a peak shape PS is generated in which the anti-resonant frequency is larger than the resonant frequency, as shown in Figure 10(c). That is, in the peak shape PS of the main lobe 90, the absolute value PL of the level difference of the positive slope line segment P from the anti-resonant frequency to the resonant frequency is larger than the absolute value ML of the level difference of the negative slope line segment M. In contrast, in the peak shape PS of the side lobe 92, the absolute value ML of the level difference of the negative slope line segment M from the resonant frequency to the anti-resonant frequency is larger than the absolute value PL of the level difference of the positive slope line segment P. Therefore, by using the normalized gradient level distribution ratio and comparing the absolute value PL of the level difference of the peak positive gradient line segment P with the absolute value ML of the level difference of the peak negative gradient line segment M, it becomes possible to reliably suppress the detection of unnecessary side lobes 92.

Furthermore, in the embodiment of the present invention, the ground beacon detection device 10 determines the presence or absence of a ground beacon 80 based on the number of peak shapes PS simultaneously extracted from the spectral waveform SW' approximated as a piecewise linear waveform by the spectral feature extraction unit 52 (see S200 in Figure 8). Here, noise from vehicle equipment and track circuit equipment generally consists mostly of pulse noise originating from the power supply system, and although short in duration, it generates broadband components. Therefore, in such cases, multiple peak shapes PS are extracted simultaneously, as shown in Figure 11. Consequently, by utilizing the number of peak shapes PS for ground beacon 80 detection, noise can be eliminated more efficiently, thereby reducing the processing load on the CPU 50 while improving the detection accuracy of the ground beacon 80.

Furthermore, in the embodiment of the present invention, the ground beacon detection device 10 estimates equipment and environmental abnormalities based on the number of peak shapes PS simultaneously extracted from the spectral waveform SW' approximated as a piecewise linear shape by the spectral feature extraction unit 52, and the number of peak shapes PS counted as candidates for ground beacon 80 from among those simultaneously extracted (see S200-S250 in Figure 8). The peak shapes PS counted as candidates for ground beacon 80 are determined using the normalized peak primary attribute group and the normalized peak secondary attribute group (see S30-S110 in Figure 7). That is, if multiple peak shapes PS are simultaneously extracted from the spectral waveform SW', there is a high probability that this is due to noise, as described above, and this can be used to estimate environmental abnormalities such as noise. Also, if there are multiple peak shapes PS counted as candidates for ground beacon 80, there is a high probability that this is an equipment abnormality in the ground beacon detection device 10, and this can be used to estimate equipment abnormalities such as failures. This makes it possible to achieve high-precision detection of ground beacon 80, as well as equipment monitoring and noise monitoring.

Moreover, unlike conventional frequency-shifting ATS systems, the ground signal detection device 10 according to the embodiment of the present invention can implement the processing from the frequency discrimination unit 40 onward shown in Figure 2 based on digital signal processing using general-purpose digital circuits and software. Furthermore, the various attributes calculated by the spectral feature extraction unit 52 include general indicators used in spectral analysis, and can therefore be used for spectral analysis and various judgments. For this reason, the normalized peak processing performed by the spectral feature extraction unit 52 is a highly versatile technique. In addition, since the amount of information is compressed by the normalized peak processing, if only the normalized peak is recorded, the amount of data can be reduced compared to spectral recording, making it possible to reduce the cost of log recording devices, etc.
Furthermore, the ground beacon detection method according to the embodiment of the present invention can be performed by the ground beacon detection device 10 according to the embodiment of the present invention, thereby achieving the same effects and advantages as the ground beacon detection device 10 according to the embodiment of the present invention.

10: Ground coil detection device, 12: On-board coil, 14: Primary coil, 16: Secondary coil, 22: Transmission processing unit, 40: Frequency discrimination unit, 52: Spectral feature extraction unit, 60: Ground coil detection unit, 80: Ground coil, 84: Track, 86: Railway vehicle, f1 to f8: Multiple predetermined frequencies, Δf: Frequency interval, SW (SW1, SW2): Spectral waveform, SW': Spectral waveform approximated as a piecewise linear shape, PS: Peak shape, P: Positive gradient line segment, M: Negative gradient line segment, F: Flat line segment, NL: Normalized peak level, NB: Normalized peak bandwidth, NF: Normalized peak frequency, NS: Normalized gradient, PL: Absolute value of the level difference of the positive gradient line segment, ML: Absolute value of the level difference of the negative gradient line segment

Claims (14)

A ground beacon detection device for detecting ground beacons installed along railway tracks, each having one of a plurality of predetermined frequencies as its resonant frequency, from a moving railway vehicle,
A vehicle-mounted coil having a primary coil and a secondary coil that are loosely coupled to each other, and which electromagnetically couples with the ground coil when it is close to the ground coil,
A transmission processing unit that injects into the primary coil a transmission wave of a discrete frequency spectrum that includes multiple frequencies encompassing all of the above-mentioned predetermined frequencies , with frequency intervals smaller than the frequency intervals of the above-mentioned predetermined frequencies,
A frequency discrimination unit that extracts the discrete frequency spectrum from the received wave received by the secondary coil,
A spectral feature extraction unit extracts a predetermined shape from a spectral waveform defined by the levels of each frequency component of the discrete frequency spectrum extracted by the frequency discrimination unit,
A ground beacon detection device characterized by including a ground beacon detection unit that detects the ground beacon according to the presence or absence of the predetermined shape. The transmission processing unit injects the transmission wave into the primary coil at a constant voltage.
The ground signal detection device according to claim 1, characterized in that the frequency discrimination unit performs level conversion on the extracted discrete frequency spectrum based on the received voltage of each frequency component. The ground coil detection device according to claim 1, characterized in that the discrete frequency spectrum has a bandwidth that includes the anti-resonant frequency of the combined impedance during electromagnetic coupling between the vehicle-mounted coil and the ground coil. The ground coil detection device according to claim 3, characterized in that the spectral feature extraction unit extracts a predetermined shape in which the level rises from the anti-resonant frequency toward the higher frequency side to the resonant frequency at the time of electromagnetic coupling between the on-board coil and the ground coil, and then the level decreases toward the higher frequency side. The ground signal detection device according to claim 1, characterized in that the spectral feature extraction unit approximates the spectral waveform in a piecewise manner by combining arbitrary line segments from three types of line segments: positive gradient line segments whose level increases toward the higher frequency side, negative gradient line segments whose level decreases toward the higher frequency side, and flat line segments that do not fall under either the positive gradient line segments or the negative gradient line segments. The ground signal detection device according to claim 5, characterized in that the spectral feature extraction unit calculates, for each line segment of the piecewise approximated spectral waveform, at least the frequency and level of the starting point, which is the endpoint on the low-frequency side, and the frequency and level of the ending point, which is the endpoint on the high-frequency side, as a primary attribute group for the line segment. The ground signal detection device according to claim 6, characterized in that the spectral feature extraction unit calculates, for each line segment of the piecewise approximated spectral waveform, at least the level difference between the starting point and the ending point, and the bandwidth (which is the frequency difference between the starting point and the ending point), as a group of quadratic line segment attributes. The ground signal detection device according to claim 7, characterized in that the spectral feature extraction unit extracts peak shapes from the piecewise approximated spectral waveform that change from a positive gradient line segment to a negative gradient line segment toward the higher frequency side, and/or change from a positive gradient line segment to a negative gradient line segment via a flat line segment with a small bandwidth toward the higher frequency side. The spectral feature extraction unit, when the positive gradient line segments constituting the peak shape are treated as peak positive gradient line segments and the negative gradient line segments constituting the peak shape are treated as peak negative gradient line segments, calculates for each of the peak shapes, at least the normalized peak level, which is the sum of the absolute value of the level difference of the peak positive gradient line segment and the absolute value of the level difference of the peak negative gradient line segment; the normalized peak bandwidth, which is the sum of the bandwidth of the peak positive gradient line segment and the bandwidth of the peak negative gradient line segment; and the normalized peak frequency, which is the average of the frequency of the endpoint of the peak positive gradient line segment and the frequency of the starting point of the peak negative gradient line segment, as a normalized peak primary attribute group.
The ground tone detection device according to claim 8, characterized in that the ground tone detection unit determines whether or not the peak shape indicates the ground tone using the normalized peak primary attribute group. The spectral feature extraction unit calculates a normalization gradient, which is the ratio of the normalized peak level to the normalized peak bandwidth, for each of the peak shapes as a group of normalized peak secondary attributes.
The ground beacon detection device according to claim 9, characterized in that the ground beacon detection unit determines whether or not the peak shape indicates the ground beacon using the normalization gradient. The spectral feature extraction unit further calculates a normalized gradient level distribution ratio for each of the peak shapes, which is the ratio of the absolute value of the level difference of the positive gradient line segment of the peak to the absolute value of the level difference of the negative gradient line segment of the peak, as the normalized peak secondary attribute group.
The ground beacon detection device according to claim 10, characterized in that the ground beacon detection unit determines whether or not the peak shape indicates the ground beacon using the normalized gradient level distribution ratio. The ground tone detection device according to claim 8, characterized in that the ground tone detection unit determines the presence or absence of the ground tone according to the number of peak shapes simultaneously extracted from the spectral waveform approximated as a piecewise linear shape. The ground tonearm detection unit is characterized by estimating the occurrence of equipment malfunctions and environmental abnormalities in the ground tonearm detection device according to claim 11, based on the number of peak shapes simultaneously extracted from the spectral waveform approximated as a piecewise linear shape, and the number of peak shapes counted as candidates for the ground tonearm, determined from among the simultaneously extracted peak shapes using the normalized primary peak attribute group and the normalized secondary peak attribute group. A ground beacon detection method for detecting ground beacons installed along railway tracks, which have one of a plurality of predetermined frequencies as their resonant frequency, from a moving railway vehicle,
A train-mounted device having a primary coil and a secondary coil that are loosely coupled to each other, and which electromagnetically couples with the ground beacon when it is close to the ground beacon, is installed on the railway vehicle.
A transmission wave with a discrete frequency spectrum that includes multiple frequencies encompassing all of the above-mentioned predetermined frequencies , with a frequency interval smaller than the frequency interval of the above-mentioned predetermined frequencies, is injected into the primary coil.
The discrete frequency spectrum is extracted from the received wave received by the secondary coil.
From the spectral waveform defined by the levels of each frequency component of the extracted discrete frequency spectrum, a predetermined shape is extracted.
A ground beacon detection method characterized by detecting the ground beacon according to the presence or absence of the predetermined shape.