JP2019191131A - Signal processing system and program - Google Patents

Abstract

To provide a signal processing system and a program with which it is possible to maintain the accuracy of identifying an object or the motion of the object irrespective of the size, shape or motion of the object.SOLUTION: A signal processing system 2 comprises a frequency analysis unit 23, a signal strength processing unit 24, an object identification unit 261, and a time setting unit 25. The signal strength processing unit 24 executes smoothing processing for smoothing signal strength and normalization processing for normalizing signal strength on a bank signal per target bank, respectively, and finds normalized strength per target bank. The object identification unit 261 performs identification processing for identifying an object or the motion of the object by the normalized strength per target bank. The time setting unit 25 is capable of variably setting the length of time of a time window. The signal strength processing unit 24 executes smoothing processing (second smoothing processing) on the bank signal each time the time window is shifted by a prescribed slide time.SELECTED DRAWING: Figure 1

Description

  The present invention generally relates to a signal processing system and a program.

  Conventionally, object detection that determines the presence or absence of a detection target in a detection area by dividing the sensor signal output from an active sensor into a plurality of frequency bands and comparing the sensor signal for each frequency band with a threshold value. An apparatus is provided (see, for example, Patent Document 1). The detection target in Patent Document 1 is a moving body such as a human body or a vehicle.

JP 2011-47779 A

  When the size, shape, or movement of an object to be detected (detection target) is predetermined to a specific size, shape, or movement, the signal processing system (object detection device) compares the objects. Can be easily identified.

  However, if the size, shape, or movement of an object is not determined in advance to a specific size, shape, or movement, the identification accuracy may be reduced depending on the size, shape, or movement of the object. It was.

  Accordingly, an object of the present invention is to provide a signal processing system and a program capable of maintaining the identification accuracy of the object or the movement of the object regardless of the size, shape, or movement of the object.

  A signal processing system according to an aspect of the present invention includes a frequency analysis unit, a signal strength processing unit, a recognition unit, and a time setting unit. The frequency analysis unit receives a sensor signal from a sensor that receives a radio signal reflected by an object, and obtains the intensity of the sensor signal in each of a plurality of filter banks as a signal intensity. The signal strength processing unit sets at least two filter banks among the plurality of filter banks as target banks. The signal strength processing unit includes a smoothing process for smoothing the signal strength within a time window applied to a bank signal representing a time change of the signal strength for each target bank, and a normalization process for normalizing the signal strength. To obtain the normalized intensity of the sensor signal for each target bank. The recognition unit performs a recognition process for identifying the object or the movement of the object based on at least one of a frequency distribution of the sensor signal determined from the normalized intensity for each target bank or a component ratio of the normalized intensity. . The time setting unit may variably set the time length of the time window. The signal strength processing unit executes the smoothing process every time the time window is shifted by a predetermined slide time with respect to the bank signal.

  A program according to an aspect of the present invention causes a computer system to function as a frequency analysis unit, a signal strength processing unit, a recognition unit, and a time setting unit. The frequency analysis unit receives a sensor signal from a sensor that receives a radio signal reflected by an object, and obtains the intensity of the sensor signal in each of a plurality of filter banks as a signal intensity. The signal strength processing unit sets at least two filter banks among the plurality of filter banks as target banks. The signal strength processing unit includes a smoothing process for smoothing the signal strength within a time window applied to a bank signal representing a time change of the signal strength for each target bank, and a normalization process for normalizing the signal strength. To obtain the normalized intensity of the sensor signal for each target bank. The recognition unit performs a recognition process for identifying the object or the movement of the object based on at least one of a frequency distribution of the sensor signal determined from the normalized intensity for each target bank or a component ratio of the normalized intensity. . The time setting unit may variably set the time length of the time window. The signal strength processing unit executes the smoothing process every time the time window is shifted by a predetermined slide time with respect to the bank signal.

  As described above, the present invention has an effect that the identification accuracy of an object can be maintained regardless of the size, shape, movement, etc. of the object.

FIG. 1 is a block diagram illustrating a detection system including a sensor and a signal processing system in the embodiment. FIG. 2A is an explanatory diagram showing a smoothing process in the signal processing system of the above. FIG. 2B is an explanatory diagram showing the first smoothing process described above. FIG. 2C is an explanatory diagram showing the second smoothing process described above. FIG. 3A is an explanatory diagram showing signal strength before normalization in the signal processing system same as above. FIG. 3B is an explanatory diagram showing signal strength after normalization. FIG. 4 is an explanatory diagram showing setting of a time window in the signal processing system same as above. FIG. 5A is a waveform diagram showing each waveform of signal intensity in the signal processing system same as above. FIG. 5B is a waveform diagram showing each waveform of the normalized intensity as described above. FIG. 6A is a waveform diagram showing a waveform of normalized intensity under the first situation in the detection area in the signal processing system same as above. FIG. 6B is a waveform diagram showing a waveform of normalized intensity under the second situation in the detection area same as above. FIG. 6C is a waveform diagram showing a waveform of normalized intensity under a third situation in the detection area same as above. 7A, 7B, and 7C are explanatory diagrams of recognition processing by principal component analysis in the signal processing system same as above. FIG. 8 is an explanatory diagram of recognition processing by multiple regression analysis in the signal processing system same as above. FIG. 9 is a waveform diagram showing the normalized strength in the signal processing system same as above. FIG. 10A is a waveform diagram of an output signal in the above signal processing system. FIG. 10B is another waveform diagram of the output signal. FIG. 11 is an explanatory diagram of normalized strength in the signal processing system same as above. FIG. 12 is another explanatory diagram of the normalized strength in the signal processing system same as above. FIG. 13 is another explanatory diagram of the normalized strength in the signal processing system same as above. FIG. 14 is another explanatory diagram of the normalized strength in the signal processing system same as above. 15A and 15B are explanatory diagrams showing signal strength in the signal processing system of the above. FIG. 16 is an explanatory diagram showing signal strength in the signal processing system of the above. FIG. 17 is an explanatory diagram showing the basic operation of the state machine included in the signal processing system of the above. FIG. 18 is an explanatory diagram showing the operation of the state machine provided in the signal processing system of the above. FIG. 19 is an explanatory diagram showing the operation of the sensor described above. 20A, 20B, and 20C are waveform diagrams showing phase difference waveforms of the signal processing system described above, and FIGS. 20A, 20B, and 20C show waveforms when the operating conditions are different from each other. FIG. 21 is a flowchart showing a schematic operation of the signal processing system of the above. FIG. 22 is a flowchart showing the second smoothing process of the signal processing system of the above.

  The following embodiments generally relate to a signal processing system and a program. More specifically, the present invention relates to a signal processing system for processing a sensor signal from a sensor that receives a radio signal (reflected wave) reflected by an object, and a program. The embodiment described below is only an example of the embodiment of the present invention. The present invention is not limited to the following embodiments, and various modifications can be made according to design and the like as long as the effects of the present invention can be achieved.

  FIG. 1 shows a configuration of a detection system 3 of the present embodiment. The detection system 3 includes a sensor 1 and a signal processing system 2. The sensor 1 transmits a radio wave (transmitted wave) in the detection area, receives a radio wave (reflected wave) reflected by an object in the detection area, and outputs a sensor signal corresponding to the movement of the object or the object. This is a sensor of the formula. The signal processing system 2 recognizes the object in the detection area or the movement of the object by performing signal processing on the sensor signal output from the sensor 1. Note that “recognizing an object or the movement of an object” means determining (identifying) whether or not an object in a detection area is an object to be detected (detection target object) or an object in the detection area. This corresponds to determining (identifying) whether or not the movement is a movement of an object to be detected (detection target event). The detection target is, for example, a person or a human hand. The detection target event is, for example, a person's movement or a person's breathing. The detection target object and the detection target event are not limited to the specific detection target object and the detection target event.

  Hereinafter, the configuration of the detection system 3 will be described in detail (see FIG. 1).

  The sensor 1 transmits a transmission wave (radio wave) from the transmission antenna, and controls the frequency of the transmission wave, the transmission timing, and the like. The transmission wave is preferably a quasi-millimeter wave of 10 GHz to 30 GHz. For example, the transmission wave is preferably a quasi-millimeter wave having a frequency value of 24.15 GHz. The transmission wave is not limited to a quasi-millimeter wave, and may be a millimeter wave or a microwave. Further, the value of the frequency of the transmission wave is not particularly limited.

  The sensor 1 receives the reflected wave reflected by the object in the detection area via the receiving antenna. The sensor 1 outputs a sensor signal corresponding to the distance from the sensor 1 to the object if the received intensity of the reflected wave is equal to or greater than a predetermined detection threshold. The sensor signal output from the sensor 1 is an analog time-domain signal corresponding to the object or the movement of the object.

  An example of the sensor 1 is a Doppler sensor. The Doppler sensor transmits a transmission wave having a predetermined frequency toward the detection area, receives a reflected wave reflected by an object in the detection area, and corresponds to a frequency difference between the transmitted transmission wave and the received reflected wave. Outputs sensor signal of Doppler frequency. When the object reflecting the radio wave is moving within the detection area, the frequency of the reflected wave is shifted by the Doppler effect.

  The signal processing system 2 includes an amplification unit 21, an A / D conversion unit 22, a frequency analysis unit 23, a signal intensity processing unit 24, a time setting unit 25, a recognition unit 26, a database 260, and an output unit 27 as main components. Furthermore, the signal processing system 2 preferably includes a misrecognition prevention processing unit 28, an area setting unit 291 and a movement identification unit 292.

  Although the signal processing system 2 of the present embodiment is configured by one signal processing device, the signal processing system 2 may be configured by a plurality of devices. For example, the signal processing system 2 may be configured by a first device, a second device, and a third device. In this case, the first device includes an amplification unit 21 and an A / D conversion unit 22. The second apparatus includes a frequency analysis unit 23, a signal strength processing unit 24, a time setting unit 25, a part of the erroneous recognition prevention processing unit 28, an area setting unit 291, and a movement identification unit 292. The third device includes a part of the recognition unit 26, the database 260, the output unit 27, and the erroneous recognition prevention processing unit 28. The first device, the second device, and the third device exchange signals with each other by wired communication (wired LAN, dedicated line communication) or wireless communication (wireless LAN, low power wireless, etc.). In the case where the signal processing system 2 is configured by a plurality of devices, the number of devices and the allocation configuration of each unit to each device are not limited to a specific form.

  The signal processing system 2 includes a computer system, and the computer system mainly includes a processor and a memory as hardware. Then, when the processor executes the program recorded in the memory, the frequency analysis unit 23, the signal strength processing unit 24, the time setting unit 25, the recognition unit 26, the misrecognition prevention processing unit 28, and the area setting unit in the present embodiment. Each function of 291 and the movement identification part 292 is implement | achieved. The program may be recorded in advance in the memory of the computer system, but may be provided through an electric communication line, or may be stored in a non-temporary recording medium such as a memory card, an optical disk, or a hard disk drive that can be read by the computer system. It may be recorded and provided. A processor of a computer system includes one or more electronic circuits including a semiconductor integrated circuit (IC) or a large scale integrated circuit (LSI). The plurality of electronic circuits may be integrated on one chip, or may be distributed on a plurality of chips. The plurality of chips may be integrated into one device, or may be distributed and provided in a plurality of devices.

  The amplifying unit 21 receives an analog time-domain sensor signal from the sensor 1 and performs amplification processing for amplifying the sensor signal. The amplifying unit 21 includes an amplifier using an operational amplifier, for example. The A / D conversion unit 22 performs A / D conversion processing for converting the analog sensor signal amplified by the amplification unit 21 into a digital sensor signal and outputting the digital sensor signal.

  The frequency analysis unit 23 performs orthogonal transform that converts the time-domain sensor signal output from the A / D conversion unit 22 into a frequency-domain sensor signal by digital signal processing. Specifically, as shown in FIG. 2A, the frequency analysis unit 23 extracts signals corresponding to each of a plurality of frequency bins 5b having different frequency bands from the sensor signal in the frequency domain. Then, the frequency analysis unit 23 sets a plurality of filter banks 5a as a filter bank group by setting a predetermined number (five in FIG. 2A) of frequency bins 5b continuous in the frequency axis direction as one filter bank 5a. . That is, the frequency analysis unit 23 extracts signals corresponding to each of the plurality of filter banks 5a having different frequency bands from the sensor signal in the frequency domain, and obtains the intensity of the sensor signal in each of the filter banks 5a as the signal intensity. Perform filter bank processing. The resolution of each filter bank 5a is determined by the width of the frequency bin 5b (Δf in FIG. 2A). The number of frequency bins 5b constituting the filter bank 5a may be other than five and is not limited to a specific number. Further, the filter bank group includes a predetermined number (for example, 16) of filter banks 5a, but the number of filter banks 5a is not limited to a specific number. Further, the number of frequency bins 5b constituting the filter bank 5a may be different for each filter bank 5a.

  The frequency analysis unit 23 of the present embodiment has a function of discrete cosine transform (DCT). Then, the frequency analysis unit 23 performs DCT processing for converting the time-domain sensor signal output from the A / D conversion unit 22 into a frequency-domain sensor signal by DCT. In this case, the frequency bin 5b of the filter bank 5a using DCT is also called a DCT bin.

  Further, the orthogonal transform executed by the frequency analysis unit 23 is not limited to DCT, and may be, for example, fast Fourier transformation (FFT). The frequency bin 5b of the filter bank 5a using FFT is also referred to as FFT bin. The orthogonal transform executed by the frequency analysis unit 23 may be a wavelet transform (WT).

  The signal strength processing unit 24 includes a smoothing unit 241 and a normalization unit 242. The signal strength processing unit 24 acquires the signal strength for each filter bank 5a from the frequency analysis unit 23, and performs smoothing processing (first smoothing processing and second smoothing processing) and standards on the acquired signal strength. Signal strength processing is performed.

  The smoothing unit 241 can execute the first smoothing process and the second smoothing process. In the first smoothing process, the signal strength of each filter bank 5a is obtained by smoothing the signal strength of each frequency bin 5b in the frequency domain (frequency axis direction) for each filter bank 5a. The second smoothing process smoothes the signal intensity in the time domain (time axis direction) for each filter bank 5a. The smoothing unit 241 can reduce the influence of noise (dark noise or ambient noise) included in the sensor signal by executing the first smoothing process and the second smoothing process.

The first smoothing process is realized by an average value filter, a load average filter, a median filter, a load median filter, or the like provided in the smoothing unit 241. As shown in FIG. 2A, for the first filter bank 5a counting from the lowest frequency, the signal strengths of the five frequency bins 5b at time t1 are s1, s2, s3, s4, and s5. . When the first smoothing process is realized by an average value filter, if the signal strength subjected to the first smoothing process is m11 (see FIG. 2B) for the first filter bank 5a, m11 is as follows. It is calculated | required by the formula 1.
m11 = (s1 + s2 + s3 + s4 + s5) / 5 (Equation 1)
Similarly, the signal intensities of the second to fifth filter banks 5a counted from the lowest frequency are m21, m31, m41, and m51, respectively, as shown in FIG. 2B. That is, in this embodiment, for convenience of explanation, the jth (j is a natural number) th filter bank that has been subjected to the first smoothing process at time ti on the time axis (i is an integer and represents a sample number). The signal strength of 5a is represented as signal strength mji.

  The second smoothing process is realized by an average value filter, a load average filter, a median filter, a load median filter, or the like provided in the smoothing unit 241. For example, when the second smoothing process is realized by an average value filter, the smoothing unit 241 includes a plurality of (for example, three) samples (for example, three) arranged in the time axis direction in each of the filter banks 5a as illustrated in FIG. 2C. Each signal intensity mji at a plurality of times is extracted by a time window W (see FIG. 4). Then, the smoothing unit 241 obtains an average value of the signal intensities mji of a plurality of samples for each of the filter banks 5a, and sets the obtained average value as the signal intensity mj after the second smoothing process in the filter bank 5a. That is, in the present embodiment, for convenience of explanation, the signal strength that has been subjected to the first smoothing process and the second smoothing process in the jth filter bank 5a is represented as mj.

For the first filter bank 5a, assuming that the signal intensity subjected to the second smoothing process is m1, m1 is obtained by the following expression 2 (see FIGS. 2C and 3A).
m1 = (m10 + m11 + m12) / 3 (2)
It becomes.

  In FIG. 2C, the second smoothing process is performed on the signal intensity m11 at time t1 (sample number i = 1). In this case, time passes in the order of time t0, t1, t2,... (Sample number i = 0, 1, 2,...). The signal intensities m10, m11, m12,... In the first filter bank 5a are sample values included in the bank signal Y1 in the first filter bank 5a. The bank signal Y1 represents a time change of the signal intensity m1i in the first filter bank 5a. That is, in this embodiment, the average of the signal strength m10 at time t0 (sample number i = 0), the signal strength m11 at time t1 (sample number i = 1), and the signal strength m12 at time t2 (sample number i = 2). The value is the result of the second smoothing process for the signal intensity m11 at time t1. The bank signal of the jth filter bank 5a is represented as Yj.

Similarly, the signal intensities m2, m3, m4, and m5 of the second to fifth filter banks 5a are obtained by the following equations 3 to 6 (see FIGS. 2C and 3A).
m2 = (m20 + m21 + m22) / 3 (Equation 3)
m3 = (m30 + m31 + m32) / 3 (Equation 4)
m4 = (m40 + m41 + m42) / 3 (Equation 5)
m5 = (m50 + m51 + m52) / 3 (Equation 6)
In this case, signal intensities m20, m21, m22,... In the second filter bank 5a are sample values included in the bank signal Y2 in the second filter bank 5a. Further, signal intensities m30, m31, m32,... In the third filter bank 5a are sample values included in the bank signal Y3 in the third filter bank 5a. Further, the signal intensities m40, m41, m42,... In the fourth filter bank 5a are sample values included in the bank signal Y4 in the fourth filter bank 5a. Further, the signal intensities m50, m51, m52,... In the fifth filter bank 5a are sample values included in the bank signal Y5 in the fifth filter bank 5a.

  The bank signal Y2 represents a time change of the signal strength m2i in the second filter bank 5a, and the bank signal Y3 represents a time change of the signal strength m3i in the third filter bank 5a. The bank signal Y4 represents a time change of the signal strength m4i in the fourth filter bank 5a, and the bank signal Y5 represents a time change of the signal strength m5i in the fifth filter bank 5a.

  It should be noted that the smoothing unit 241 according to the present embodiment described above applies only to at least two filter banks 5a used by the object recognition unit 261 (described later) included in the recognition unit 26 for recognition processing among all the filter banks 5a. A first smoothing process and a second smoothing process are performed. Hereinafter, among all the filter banks 5a, the filter bank 5a used by the object recognition unit 261 for the recognition process is referred to as a target bank 5a. That is, the smoothing unit 241 performs the first smoothing process and the second smoothing process described above only on the target bank 5a among all the filter banks 5a. The target bank 5a may be a part of all the filter banks 5a, or all the filter banks 5a may be the target bank 5a.

  The normalization unit 242 normalizes the signal strength mj of each target bank 5a with the sum of the signal strengths mj of the target bank 5a, and obtains the normalized strength nj of each target bank 5a. In the present embodiment, the normalized strength of the jth filter bank 5a is represented as nj.

For example, it is assumed that the total number of the filter banks 5a is 16, and the target banks 5a are only the 1st to 5th, counting in order from the lowest frequency. When the signal strength m1 (see FIG. 3A) of the first target bank 5a at time t1 (sample number i = 1) is normalized, the normalized strength n1 is obtained by the following Expression 7 (see FIG. 3B). .
n1 = m1 / (m1 + m2 + m3 + m4 + m5) (Equation 7)
Similarly, when the signal strengths m2-m5 (see FIG. 3A) of the second to fifth target banks 5a are normalized, the normalized strengths n2, n3, n4, and n5 are expressed by the following equations 8- 11 (see FIG. 3B).
n2 = m2 / (m1 + m2 + m3 + m4 + m5) (Equation 8)
n3 = m3 / (m1 + m2 + m3 + m4 + m5) (Equation 9)
n4 = m4 / (m1 + m2 + m3 + m4 + m5) (Equation 10)
n5 = m5 / (m1 + m2 + m3 + m4 + m5) (Equation 11)
Further, when the signal strength mj of each target bank 5a is normalized by the sum of the signal strengths mj of the target bank 5a, the normalization unit 242 may use logarithmic values of the signal strengths mj. In this case, the normalization unit 242 performs logarithmic conversion that converts each signal intensity mj from a linear scale value to a logarithmic scale value (logarithmic value). Then, the normalization unit 242 performs normalization processing by subtraction processing on a logarithmic scale that subtracts the logarithmic value of the sum of the signal strengths mj of the target bank 5a from the logarithmic value of the signal strength mj of each target bank 5a. Can do. On the other hand, when the normalization process is performed on a linear scale, a division process for dividing the signal strength mj of each target bank 5a by the sum of the signal strengths mj of the target bank 5a is required. That is, the normalization unit 242 can reduce the calculation load required for the normalization process by using the logarithmic value of each signal strength mj for the normalization process.

  Note that the normalization unit 242 may normalize the signal strength mj of each target bank 5a with the sum of the signal strengths mj of all the filter banks 5a, and obtain the normalized strength nj of each target bank 5a. .

  The smoothing process of the smoothing unit 241 and the normalization process of the normalizing unit 242 may be performed after the smoothing process as described above, or smoothed after the normalization process. Processing may be executed. That is, the smoothing unit 241 has a function of smoothing the signal strength mji of each filter bank 5a, and the normalization unit 242 has a function of normalizing the signal strength mji of each filter bank 5a.

  The recognition unit 26 includes an object recognition unit 261 and a background recognition unit 262.

  Then, the object recognition unit 261 detects the detection target in the detection area or at least one of the frequency distribution of the sensor signal determined from the normalized strength nj for each target bank 5a and the component ratio of the sensor signal determined from the normalized strength nj. A recognition process for identifying the detection target event is performed.

  When the object in the detection area is a detection target, the frequency distribution of the sensor signal and the component ratio of the sensor signal determined from the normalized intensity nj are the frequency distribution and sensor when the object in the detection area is other than the detection target. The signal component ratios tend to be different from each other. Further, when the movement of the object in the detection area is a detection target event, the frequency distribution of the sensor signal and the component ratio of the sensor signal determined from the normalized intensity nj are other than the detection target event. In this case, the frequency distribution and the sensor signal component ratio tend to be different. Therefore, the object recognition unit 261 performs recognition processing for identifying the detection target object or the detection target event based on the frequency distribution or the component ratio of the sensor signal determined from the normalized strength nj in each target bank 5a. And the signal processing system 2 of this embodiment has reduced the influence of the noise (dark noise or the noise which generate | occur | produces around) which is contained in a sensor signal by performing a smoothing process and a normalization process, and is high Recognition accuracy can be obtained. Here, recognition is a concept including identification, classification, detection, and determination.

  The frequency distribution and component ratio of the sensor signal input to the object recognizing unit 261 is determined by the normalized intensity nj obtained by smoothing and normalizing the signal that has passed through the predetermined plurality of filter banks 5a. The frequency distribution and component ratio of the sensor signal are statistically different from each other for each of a plurality of different objects or for each of a plurality of different object movements. Then, by using the normalized intensity nj smoothed and standardized by the signal processing system 2, a statistical difference in frequency distribution and component ratio of sensor signals with respect to movements of different objects or different objects. Is further emphasized. Therefore, the signal processing system 2 of the present embodiment can reduce misrecognition caused by an object other than the detection target and misrecognition caused by the movement of the object other than the detection target. As a result, erroneous recognition can be reduced.

  Further, the signal processing system 2 includes a time setting unit 25. The time setting unit 25 sets the value of the time length of the time window W (see FIG. 4) used in the second smoothing process described above. In FIG. 4, in order to distinguish a plurality of time windows W, symbols W1, W2, W3,.

  The smoothing unit 241 extracts each signal intensity mji of a plurality of samples in the time axis direction in each of the filter banks 5a. Then, the smoothing unit 241 obtains an average value of the signal intensities mji of a plurality of samples for each of the filter banks 5a, and sets the obtained average value as the signal intensity mj after the second smoothing process in the filter bank 5a. At this time, the smoothing unit 241 extracts the signal intensities mji of a plurality of samples from the bank signals Y1-Y5 by applying the time windows W to the bank signals Y1-Y5, respectively.

  The time setting unit 25 can set the time length of the time window W to be variable by setting the value of the time length of the time window W to an arbitrary value as the window setting value Ta (see FIG. 4). it can. When the time setting unit 25 increases the window setting value Ta, the time length of the time window W is increased, and the number of samples of the signal intensity mji extracted for one second smoothing process is increased. When the time setting unit 25 decreases the window setting value Ta, the time length of the time window W is shortened and the number of samples of the signal intensity mji extracted for one second smoothing process is decreased. .

  The time setting unit 25 includes at least an automatic setting function for automatically setting the window setting value Ta according to a detection object or a detection target event, and a manual setting function for setting the window setting value Ta to a value instructed by a user operation. Have one.

  The automatic setting function of the time setting unit 25 performs, for example, the following setting process.

  First, the operation mode of the signal processing system 2 is set to the data collection mode. The frequency analysis unit 23 acquires a time-domain sensor signal and converts it to a frequency-domain sensor signal when there is a detection target object or detection target event in the detection area. And the time setting part 25 analyzes the frequency component contained in the sensor signal corresponding to a detection target object or a detection target event based on the sensor signal of a frequency domain. The time setting unit 25 reduces the window setting value Ta when the dominant frequency component among the frequency components included in the sensor signal is relatively high. Further, the time setting unit 25 relatively increases the window setting value Ta when the dominant frequency component among the frequency components included in the sensor signal is relatively low. That is, the time setting unit 25 decreases the window setting value Ta as the frequency of the dominant frequency component included in the sensor signal corresponding to the detection target object or the detection target event becomes higher (the shorter the period), the more dominant The window set value Ta is increased as the frequency of the frequency component is lower (the period is shorter). Then, after the time setting unit 25 sets the window setting value Ta, the operation mode of the signal processing system 2 shifts from the data collection mode to the detection mode, and the above-described recognition processing is performed.

  Specifically, the window setting value Ta is set so that one period or half period of the dominant signal component included in the sensor signal corresponding to the detection target object or the detection target event is included in the time window W. In this case, the window setting value Ta is preferably set to the same value (or substantially the same) as one period or half period of the signal component. Further, the window setting value Ta is preferably not more than twice the one period or half period of the signal component.

  The manual setting function of the time setting unit 25 receives an instruction value by a user operation on an operation unit such as a keyboard, a mouse, a button, or a lever, and sets the instruction value as a window setting value Ta. In this case, the user determines the instruction value based on the frequency analysis result of the sensor signal corresponding to the detection target event.

  For example, in a space where an infant exists such as a kindergarten, a nursery school, a day nursery, and a childcare facility, it is necessary to more surely check the presence of the infant so that the infant does not come off or become unclear. Therefore, the signal processing system 2 identifies whether or not an infant as a detection target exists in the detection area by identifying the infant's breathing in the detection area as the detection target event as the detection target event. . In general, in breathing of an infant, an inspiration period and an expiration period are alternately generated in a cycle of 3 seconds. In this case, it is preferable that the time setting unit 25 sets the window setting value Ta to 3 seconds (or about 3 seconds). Further, since the movement of the thorax due to the infant's breathing is about several millimeters, the movement of the thorax due to the infant's breathing can be identified by setting the transmission wave to a quasi-millimeter wave or a millimeter wave.

  Then, as illustrated in FIG. 4, the smoothing unit 241 applies the time window W to the bank signal Yj of the j-th filter bank 5a, and extracts the signal intensities mji of a plurality of samples that fall within the time window W, respectively. The smoothing unit 241 performs the above-described second smoothing process using the extracted signal strengths mji of a plurality of samples. Every time the second smoothing process is executed, the smoothing unit 241 shifts the time window W applied to the bank signal Yj by a predetermined slide time Ts in the time progress direction. That is, the smoothing unit 241 executes the second smoothing process every time the time window in which the window setting value Ta is set is shifted in the time progress direction by the slide time Ts. In FIG. 4, time windows W1, W2, and W3 are applied in the order of time windows W1, W2, and W3, respectively.

  Here, the slide time Ts is preferably shorter than the window setting value Ta. In the present embodiment, the time setting unit 25 automatically sets the slide time Ts shorter than the window setting value Ta, and the time setting unit 25 sets the value of the slide time Ts to, for example, 1/3 of the window setting value Ta. Set to the value of. When the window setting value Ta is 3 seconds, the value of the slide time Ts is set to 1 second. In this case, a part of the time window before shifting in the time progression direction (for example, W1) and the time window after shifting in the time progression direction (for example, W2) overlap in the time axis direction. Therefore, the feature of the sensor signal is extracted more reliably.

  Since the second smoothing process is executed so that the detection target object or the detection target event can be accurately identified by the setting process of the window setting value Ta and the slide time Ts by the time setting unit 25 described above, object recognition is performed. The identification accuracy of the part 261 is improved. Further, since the smoothing unit 241 executes the second smoothing process every time the time window is shifted in the direction of time movement by the slide time Ts, the value of the slide time Ts is appropriately set, so that the object recognition unit The identification accuracy of H.261 can be further improved.

  FIG. 5A shows each waveform Y11-Y14 in the time domain of the signal strength mji. A waveform Y11 indicates the signal intensity mji of the sensor signal generated by the infant's inspiratory motion in the filter bank 5a of 0.3-1 Hz. A waveform Y12 indicates the signal intensity mji of the disturbance component included in the sensor signal generated by the infant's inspiratory motion in the 1-10 Hz filter bank 5a. A waveform Y13 indicates the signal intensity mji of the sensor signal generated by the infant's exhalation movement in the filter bank 5a of 0.3-1 Hz. A waveform Y14 indicates the signal intensity mji of the disturbance component included in the sensor signal generated by the infant's expiratory movement in the 1-10 Hz filter bank 5a.

  FIG. 5B shows each waveform Y21-Y24 in the time domain of the normalized strength nj. A waveform Y21 indicates the normalized intensity nj of the sensor signal generated by the infant's inspiratory motion in the 0.3-1 Hz filter bank 5a. A waveform Y22 indicates the normalized intensity nj of the disturbance component included in the sensor signal generated by the infant's inspiratory motion in the 1-10 Hz filter bank 5a. A waveform Y23 indicates the normalized intensity nj of the sensor signal generated by the infant's exhalation exercise in the 0.3-1 Hz filter bank 5a. A waveform Y24 indicates the normalized intensity nj of the disturbance component included in the sensor signal generated by the infant's exhalation movement in the 1-10 Hz filter bank 5a.

  Each waveform Y21-Y24 of the normalized intensity nj in FIG. 5B is subjected to the first smoothing process, the second smoothing process, and the normalization process, respectively, with respect to each waveform Y11-Y14 of the signal intensity mji in FIG. 5A. It is the waveform in the performed time domain.

  Next, the recognition process of the object recognition unit 261 will be described in detail.

  The object recognizing unit 261 determines the movement of the object or the object in the detection area based on at least one of the frequency distribution of the sensor signal determined from the normalized intensity nj for each target bank 5a and the component ratio of the sensor signal determined from the normalized intensity nj. A recognition process for identifying is performed.

  For example, the object recognition unit 261 identifies an object or a motion of the object in the detection area based on the characteristics of the normalized strength nj for each target bank 5a included in a predetermined period. Since the characteristic of the normalized strength nj for each target bank 5a depends on the frequency distribution of the sensor signal, this identification method corresponds to recognition processing based on the frequency distribution of the sensor signal. 6A, 6B, and 6C respectively show waveforms Y31, Y32, and Y33 in the time domain of the normalized intensity nj under different conditions in the detection area. FIG. 6A shows a waveform Y31 of the normalized intensity nj under a situation where a person in the detection area stands or sits (repeat standing and sitting alternately). FIG. 6B shows a waveform Y32 of the normalized intensity nj under a situation where a person in the detection area is sitting and standing still. FIG. 6C shows a waveform Y33 of the normalized intensity nj under the situation where no person is present in the detection area. 6A, 6B, and 6C are expressed as absolute values of normalized values.

  The database 260 stores data of threshold values R1 and R2 for the normalized strength nj in advance. The threshold value R1 is larger than the threshold value R2, and the threshold value R1 is a threshold value for identifying an operation in which a person in the detection area repeats standing and sitting alternately. On the other hand, the threshold value R2 is a threshold value for identifying a state where a person in the detection area is sitting. In this case, the characteristic of the normalized strength nj is the number of times that the waveforms Y31, Y32, Y33 have changed across the threshold value R1 or the threshold value R2 (the number of times that the waveform has increased or decreased across the threshold value).

  First, the object recognition unit 261 performs recognition processing using the threshold value R1. In this case, the object recognizing unit 261 counts each number of times that the waveform of the normalized intensity nj crosses the threshold value R1 for each predetermined period Tr that is a period of a predetermined time length. If the number of times counted with respect to the threshold value R1 is equal to or greater than a predetermined number (see FIG. 6A), the object recognition unit 261 performs an operation in which a person in the detection area repeats standing and sitting alternately. It is determined that

  If the number of times counted with respect to the threshold value R1 is less than a predetermined number, the object recognition unit 261 next performs a recognition process using the threshold value R2. In this case, the object recognition unit 261 counts the number of times that the waveform of the normalized strength nj crosses the threshold value R2 for each predetermined period Tr. Then, if the number of times counted with respect to the threshold value R2 is equal to or greater than a predetermined number (see FIG. 6B), the object recognition unit 261 determines that a person in the detection area is sitting still.

  The object recognition unit 261 determines that there is no person in the detection area when the number of times counted with respect to the threshold value R2 is less than a predetermined number (see FIG. 6C).

  Further, the object recognizing unit 261 may determine whether or not a specific frequency component is included in a waveform representing a temporal change in the normalized strength nj as a feature of the normalized strength. That is, the object recognizing unit 261 determines whether or not the standardized intensity nj changes at a specific period, and detects an object in the detection area when the standardized intensity nj changes at a specific period. It determines with it being a target object, or it determines with the motion of the object in a detection area being a detection target event. When the object in the detection area is a detection target, or when the movement of the object in the detection area is a detection target event, the sensor signal includes a specific frequency component corresponding to the detection target or the detection target event. It is. Therefore, the object recognizing unit 261 can determine whether or not there is a detection target object or a detection target event by determining whether or not a specific frequency component is included in the waveform representing the temporal change in the normalized strength nj. The specific frequency component may be one or plural. Further, “a waveform includes a specific frequency component” includes “a waveform includes a specific range of frequency components”.

  Further, the object recognition unit 261 may identify an object by performing pattern recognition processing based on principal component analysis, for example. In this case, the object recognition unit 261 operates according to a recognition algorithm using principal component analysis. When the object recognition unit 261 performs pattern recognition processing by principal component analysis, learning sample data is acquired in advance, and data obtained by performing principal component analysis on the learning sample data is stored in the database 260. Learning sample data includes learning sample data when there is no detection object or detection target event in the detection area, learning sample data corresponding to various movements (detection target events) of the detection target in the detection area, and within the detection area Learning sample data corresponding to various movements of an object other than the object to be detected. Data stored in the database 260 is data used for pattern recognition, and is category data in which an object motion is associated with a projection vector and a discrimination boundary value (threshold value).

  Here, for convenience of explanation, FIG. 7A shows respective distributions in the frequency domain of the normalized intensities n1-n5 corresponding to the learning sample data when there is no detection target object or detection target event in the detection area of the sensor 1. Moreover, each distribution in the frequency domain of the normalization intensity | strength n1-n5 corresponding to learning sample data when there exists a detection target object or a detection target event is shown to FIG. 7B. In FIG. 7A, the normalized intensities n1-n5 of each of the target banks 5a are represented by n10, n20, n30, n40, n50 in order from the low frequency side. In FIG. 7B, the normalized intensities n1 to n5 of the target banks 5a are represented by n11, n21, n31, n41, and n51 in order from the low frequency side. 7A and 7B, the sum of the normalized strengths of the three target banks 5a on the low frequency side is a variable na, and the sum of the normalized strengths of the two target banks 5a on the high frequency side is the variable. Let nb.

  In short, in FIG. 7A, na = n10 + n20 + n30 and nb = n40 + n50. In FIG. 7B, na = n11 + n21 + n31 and nb = n41 + n51.

  FIG. 7C is a two-dimensional diagram in which two variables na and nb are used as coordinate axes orthogonal to each other in order to explain the two-dimensional scatter diagram, the projection axis, and the identification boundary conceptually. In the principal component analysis, the identification boundary is determined in advance using the two data groups Gr0 and Gr1. The data group Gr0 is a data group corresponding to learning sample data when there is no detection target object or detection target event in the detection area of the sensor 1. The data group Gr1 is a data group corresponding to learning sample data when there is a detection target object or a detection target event in the detection area of the sensor 1. In FIG. 7C, the coordinate position of each scatter point (“+” in FIG. 7C) in the data group Gr0 is μ0 (nb, na), and the coordinate position of each scatter point in the data group Gr1 is μ1 (nb, na). It is said. In the principal component analysis, the distribution of data (schematically indicated by a broken line and a solid line) in which each scattered point in each region surrounded by a broken line (data group Gr0) and a solid line (data group Gr1) in FIG. 7C is projected on the projection axis. The projection axis is determined under the condition that the interval between the average values is maximized and the variance is maximized. Thereby, in principal component analysis, a projection vector can be obtained for each learning sample.

  And the object recognition part 261 calculates | requires a scatter point from the normalization intensity | strength n1-n5 in each object bank 5a, and if the projection point of the calculated | required scatter point is in one side from an identification boundary on a projection axis, it will be in a detection area. Recognize that there is an object to be detected or an event to be detected. The object recognizing unit 261 recognizes that there is no detection target object or detection target event in the detection area if the projection point of the obtained scattering point is on the other side of the identification boundary on the projection axis.

  The signal processing system 2 includes an output unit 27 that outputs a recognition result obtained by the object recognition unit 261. When the object recognition unit 261 recognizes the detection target object or the detection target event, the output unit 27 outputs an output signal “1” indicating the detection state. Further, when the object recognition unit 261 does not recognize the detection target object or the detection target event, the output unit 27 outputs an output signal “0” indicating that it is in the non-detection state.

  In addition, the object recognition unit 261 is not limited to identifying an object by pattern recognition processing by principal component analysis, and may identify an object by pattern recognition processing by KL conversion, for example. In the signal processing system 2, the object recognition unit 261 performs pattern recognition processing by principal component analysis or pattern recognition processing by KL conversion, thereby reducing the amount of calculation in the object recognition unit 261 and reducing the capacity of the database 260. Can be achieved.

  The object recognition unit 261 may perform a recognition process for identifying an object based on a component ratio of sensor signals determined from the normalized strength nj for each target bank 5a. For example, the object recognition unit 261 identifies an object by performing recognition processing based on multiple regression analysis. In this case, the object recognition unit 261 operates according to a recognition algorithm using multiple regression analysis.

  When the object recognition unit 261 performs recognition processing by multiple regression analysis, a plurality of learning data is acquired in advance, and data obtained by performing multiple regression analysis on the plurality of learning data is stored in the database 260 in advance. Keep it. The plurality of learning data is learning data corresponding to the detection target object or the detection target event in the detection area.

  According to the multiple regression analysis, as shown in FIG. 8, when the combined waveform P0 is a combined waveform of the signal component p1, the signal component p2, and the signal component p3, the types of the signal components p1, p2, and p3, Even if the numbers and the intensities of the signal components p1, p2, and p3 are unknown, the signal components p1, p2, and p3 can be estimated separately from the combined waveform P0.

Here, a matrix having signal components p1, p2, and p3 as matrix elements is [S], an inverse matrix of [S] is [S] ^-1, and coefficients are Q (q1, q2, q3). The relationship between the composite waveform P0, the matrix [S], the inverse matrix [S], and the coefficient Q is expressed by the following expressions 12 and 13. The data stored in the database 260 is data used for recognition processing, and is data in which the motion of the object is associated with the signal components p1, p2, and p3.
P0 = [S] Q (Equation 12)
Q = [S] ⁻¹ P0 (Equation 13)
As described above, the signal processing system 2 performs signal processing of the sensor signal in the frequency domain. Specifically, the object recognition unit 261 recognizes a detection target or a detection target event based on at least one of the characteristics of the frequency distribution of the sensor signal and the component ratio of the normalized intensity of the signal in the frequency domain. I do. For signal processing of the sensor signal in the frequency domain, for example, recognition processing using principal component analysis, KL conversion, multiple regression analysis, neural network, or the like is used.

  Hereinafter, in the signal processing system 2 that performs such recognition processing, a configuration that suppresses erroneous recognition due to movement of an object other than the detection target object or an object other than the detection target event will be described. In particular, the movement of an object other than the detection target object or an object other than the detection target event may cause a steep temporal fluctuation in the frequency domain signal, and erroneous recognition due to such a temporal fluctuation of the signal is suppressed. The

  First, it is assumed that the object recognition unit 261 performs recognition processing based on the component ratio of the sensor signal determined from the normalized intensity for each target bank 5a. In FIG. 9, the horizontal axis represents time, the vertical axis represents the normalized intensity nj, and the signal components A 1 and A 2 separated by multiple regression analysis from the data on the time axis of the normalized intensity nj output from the normalization unit 242. A2 is shown. Then, the object recognizing unit 261 uses a signal component A1 resulting from the movement of the detection target object (for example, movement of a person) in one or a plurality of predetermined target banks 5a as a signal resulting from the movement of an object other than the detection target object. When it is larger than the component A2, it is identified that the detection target exists. Hereinafter, the signal component A1 resulting from the movement of the detection target object is referred to as a detected object signal component A1, and the signal component A2 resulting from the movement of an object other than the detection target object is referred to as a disturbance signal component A2.

  However, the detected object signal component A1 may have a steep temporal variation due to the movement of an object other than the detection target. If there is a steep temporal variation in the detected object signal component A1, even if the first smoothing process and the second smoothing process are performed, the object recognizing unit 261 indicates the movement of an object other than the detection target object. There is a possibility of erroneous recognition (false detection) as the movement of the detection object. Therefore, the misrecognition prevention processing unit 28 includes a signal component extraction unit 281, a disturbance determination unit 282, a background signal removal unit 283, a level setting unit 284, and a parameter adjustment unit 285, thereby preventing erroneous recognition of the object recognition unit 261. Performs misrecognition prevention processing for reduction.

  First, the signal component extraction unit 281 and the disturbance determination unit 282 constitute a first erroneous recognition prevention processing unit (see FIG. 1). The signal component extraction unit 281 extracts the detected object signal component A1 from the data on the time axis of the normalized strength nj for each filter bank 5a output from the normalization unit 242. The disturbance determination unit 282 determines that the object recognition unit is in the case where the magnitude of variation per unit time of the extracted detected object signal component A1 is equal to or greater than a predetermined value (E11) (first predetermined value) in each target bank 5a. The recognition process by H.261 is prohibited, or the result of the recognition process by the object recognition unit 261 is invalidated.

  Specifically, the signal component extraction unit 281 extracts the detected object signal component A1 from the data on the time axis of the normalized strength nj for each target bank 5a using multiple regression analysis. Then, the signal component extraction unit 281 outputs the data (see FIG. 9) of the detected object signal component A1 to the disturbance determination unit 282.

  The disturbance determination unit 282 differentiates the detected object signal component A1 from the distribution of the detected object signal component A1 in the time axis region, and calculates the time differential intensity (= dA1 / dt) of the detected object signal component A1 at a predetermined time interval. calculate. Then, the disturbance determination unit 282 prohibits the recognition processing by the object recognition unit 261 when the time differential intensity of the detected object signal component A1 is equal to or greater than a predetermined value (E11) in each target bank 5a, or the object recognition unit The result of recognition processing by H.261 is invalidated.

  At this time, the target bank 5a that is the target of the recognition process availability determination using the signal component extraction unit 281 and the disturbance determination unit 282 is appropriately set according to the detection target object or the detection target event. In all the target banks 5a, or in one or more target banks 5a, when the magnitude of variation per unit time of the detected object signal component A1 is equal to or greater than a predetermined value (E11), The result of the recognition process becomes invalid.

  FIG. 10A shows an output signal output from the output unit 27 when the recognition object signal component A1 shown in FIG. 9 is not subjected to recognition processing using the signal component extraction unit 281 and the disturbance determination unit 282. It is. FIG. 10B shows an output signal output from the output unit 27 when the detection object signal component A1 shown in FIG. 9 is subjected to recognition processing using the signal component extraction unit 281 and the disturbance determination unit 282. is there. The detected object signal component A1 shown in FIG. 9 has a steep temporal variation caused by the movement of an object other than the detection target. However, it is confirmed that it is possible to suppress erroneous recognition of the movement of an object other than the detection target object as the movement of the detection target object by determining whether the recognition process using the signal component extraction unit 281 and the disturbance determination unit 282 is possible. It was.

  Therefore, the signal processing system 2 accurately detects the movement of the detection target object by the frequency domain signal analysis regardless of the presence or absence of the steep temporal fluctuation of the detection object signal component A1 caused by the movement of the object other than the detection target object. It can be detected. Further, the time required for the process of comparing the time differential intensity of the detected object signal component A1 with the predetermined value (E11) is relatively short, and the signal processing system 2 having this configuration is realized with a simple and small structure. can do.

  Further, the disturbance determination unit 282 prohibits the recognition processing by the object recognition unit 261 until the predetermined time elapses after the time differential intensity of the detected object signal component A1 becomes equal to or greater than a predetermined value (E11), or object recognition. The result of the recognition process by the unit 261 may be invalidated. For example, it is assumed that the time differential intensity of the detected object signal component A1 becomes equal to or greater than a predetermined value (E11) and then becomes a flat waveform (the time differential intensity is less than the predetermined value (E11)). In this case, the recognition process by the object recognition unit 261 is prohibited for a predetermined time after the time differential intensity of the detected object signal component A1 becomes equal to or greater than the predetermined value (E11), or the recognition process by the object recognition unit 261 is performed. Invalidate the result. Accordingly, it is possible to more reliably reduce erroneous recognition due to a sharp temporal variation of the detected object signal component A1 due to the movement of an object other than the detection target.

  Further, the disturbance determination unit 282 prohibits the recognition process by the object recognition unit 261 or the result of the recognition process by the object recognition unit 261 when the time differential intensity of the detected object signal component A1 is equal to or less than a predetermined value (E12). It may be invalid. The predetermined value (E12) used here is a value smaller than the predetermined value (E11), and is for suppressing erroneous recognition due to the detected object signal component A1 having a substantially constant signal intensity with small fluctuation.

  Further, the object recognizing unit 261 prohibits the recognition processing by the object recognizing unit 261 until a predetermined time elapses after the time differential intensity of the detected object signal component A1 becomes equal to or less than a predetermined value (E12). The result of the recognition processing by H.261 may be invalidated. In this case, erroneous recognition due to the detected object signal component A1 having a substantially constant signal intensity with small fluctuations can be more reliably suppressed.

  The disturbance determination unit prohibits the recognition processing by the object recognition unit 261 when the magnitude of fluctuation per unit time of the signal strength mji is equal to or greater than a predetermined value based on the data on the time axis of the signal strength mji. Or the result of recognition processing by the object recognition unit 261 may be invalidated.

  Further, the background signal removal unit 283 constitutes a second erroneous recognition prevention processing unit (see FIG. 1). The background signal removal unit 283 performs background signal (that is, noise) included in each signal of the filter bank 5a (data on the time axis of the signal strength mji or data on the time axis of the normalized strength nj). Then, it is removed from each signal of the filter bank 5a.

  The signal processing system 2 includes, for example, a first mode in which the background signal removal unit 283 estimates a background signal, a second mode in which recognition processing by the object recognition unit 261 is performed, and a background recognition unit 262 as a stationary operation mode. And a third mode for identifying the background signal. Then, the signal processing system 2 switches between the first mode and the second mode every predetermined time (for example, 30 seconds) timed by a timer provided in the signal processing system 2, and performs object recognition in the second mode. It is preferable to operate in the third mode when the object or the movement of the object is identified by the unit 261. Here, in the signal processing system 2, the background signal removal unit 283 estimates the background signal during the first mode period, and during the second mode period, the background signal removal unit 283 removes the background signal from the sensor signal. The object recognition unit 261 preferably performs recognition processing. The time in the first mode and the time in the second mode are not limited to the same time (for example, 30 seconds), and may be different times.

  The background signal removal unit 283 may remove the background signal from the sensor signal by subtracting the signal strength of the background signal from the signal strength (mj or nj) of each filter bank 5a. In this case, the background signal removal unit 283 can be configured by a subtracter that subtracts the estimated signal strength of the background signal from the signal strength of each filter bank 5a.

  The background signal removal unit 283 estimates the signal strength of the signal obtained for each filter bank 5a during the first mode, and updates the background signal data as needed. You may make it do. Further, in the first mode, the background signal removal unit 283 estimates the average value of the signal strengths of the plurality of signals obtained for each of the filter banks 5a as the signal strength of the background signal for each filter bank 5a. May be. That is, the background signal removing unit 283 may use an average value on the time axis of a plurality of sample signals for each filter bank 5a obtained in advance as the background signal. Thereby, the background signal removal unit 283 can improve the estimation accuracy of the background signal.

  Further, the background signal removal unit 283 may use the signal immediately before each filter bank 5a as the background signal in the first mode. Here, the background signal removal unit 283 may use the signal strength mj immediately before on the time axis as the signal strength of the background signal before the normalization unit 242 performs the normalization process on the signal strength mj. In short, the background signal removal unit 283 subtracts the signal strength mj of one sample before on the time axis from the signal strength mj to be normalized, with respect to the signal strength mj of each filter bank 5a, from the sensor signal. You may make it have a function which removes a background signal.

  By the way, depending on the surrounding environment, the frequency bin 5b including a background signal (noise) having a relatively large signal strength may be known in advance. For example, when there is a device that is supplied with power from a commercial power source around the detection system 3, the frequency is lower than the frequency of the commercial power source (for example, 60 Hz) (for example, 60 Hz, 120 Hz, etc.) The signal of the frequency bin 5b including a specific frequency is likely to include a background signal having a relatively large signal strength. On the other hand, the sensor signal output from the sensor 1 when the object is moving in the detection area has a frequency (Doppler frequency) of the sensor signal, a distance between the sensor 1 and the object, and a moving speed of the object. Therefore, a large signal intensity does not constantly occur at a specific frequency.

  Therefore, when each filter bank 5a has a plurality of frequency bins 5b, the signal processing system 2 sets the frequency bin 5b in which the background signal is constantly included as the specific frequency bin 5b. The background signal removal unit 283 invalidates the signal strength of the specific frequency bin 5b in the first mode. Then, the background signal removing unit 283 removes the background signal by using the signal strength estimated from the strengths of the signals of the two frequency bins 5b adjacent to the specific frequency bin 5b as the signal strength of the specific frequency bin 5b. To do. Thereby, the signal processing system 2 can reduce the influence of a background signal (noise) of a specific frequency that is constantly generated in a shorter time. Therefore, the signal processing system 2 can improve the detection accuracy of the detection target object or the detection target event.

  Furthermore, in the signal processing system 2, when the object or the movement of the object is identified by the object recognition unit 261, the background recognition unit 262 identifies a steady background signal and updates (changes / cancels) the background signal data. It is preferable. Further, the background recognition unit 262 may periodically identify a steady background signal and periodically update (change / cancel) the data of the background signal. Thereby, the signal processing system 2 suppresses the background signal removing unit 283 from using the previous background signal when the intensity of the stationary background signal changes or when the stationary background signal disappears. This makes it possible to further reduce misrecognition.

  The background recognition unit 262 identifies the background signal based on at least one of the characteristics of the frequency distribution determined from the normalized strength nj for each target bank 5a and the characteristics of temporal continuity of the frequency distribution.

  The characteristic of the frequency distribution is an instantaneous frequency distribution, and means, for example, a distribution in the frequency region (on the frequency axis) of the normalized strength nj at a certain time as shown in FIG. The temporal continuity of the frequency distribution means stationarity and means statistical stability.

  When the background recognition unit 262 identifies a background signal based on the characteristics of the frequency distribution, the background recognition unit 262 performs a component analysis of the characteristics of the frequency distribution to identify the background signal. In the component analysis, whether or not the characteristics of the frequency distribution are realized by any one or a combination of a plurality of frequency distribution components that are stored in the memory as a reference of the background signal, and the background signal is identified based on the result. To do. In this case, the background recognition unit 262 may perform a recognition process by multiple regression analysis, for example. The frequency distribution component means a distribution in the frequency domain of the normalized strength nj.

  For example, when the background recognizing unit 262 identifies a background signal based on the temporal continuity characteristics of the frequency distribution, the pulse recognition signal (normalized intensity) where the normalized intensity nj exceeds a specified value (for example, 0.1). nj (time waveform) is equal to or more than a specified number of times (for example, 10 times) within a predetermined time (for example, about 3 seconds) in the same filter bank 5a, the average value of the pulse-like signal for each filter bank 5a Alternatively, the average value multiplied by a constant is recognized as a stationary background signal. FIG. 12 shows an example of the frequency distribution in the case where the pulsed signal is less than the specified number of times in a predetermined time. FIG. 13 shows an example of the frequency distribution in the case where the number of pulsed signals is equal to or greater than the specified number of times in a predetermined time. FIG. 14 shows an example of an instantaneous frequency distribution including a stationary background signal.

  In addition, when the background recognition unit 262 identifies the background signal based on the temporal continuity characteristics of the frequency distribution, the flat frequency at which the normalized strength nj of each filter bank 5a exceeds a specified value (for example, 0.1). When the distribution is a predetermined number of times (for example, 10 times) or more during a predetermined time (for example, about 3 seconds), the average value of the normalized strength nj for each filter bank 5a or a constant for the average value The multiplied signal is recognized as a steady background signal.

  Furthermore, the signal processing system 2 performs the recognition process by the object recognition unit 261 or the object recognition unit 261 only when the sum of the signal strengths mj of the target bank 5a before normalization by the normalization unit 242 is greater than or equal to the threshold value. The recognition result by may be validated. FIGS. 15A and 15B are examples in which the signal strength mj of each of the target banks 5a before normalization by the normalization unit 242 is set to m1, m2, m3, m4, and m5 in order from the low frequency side. FIG. 15A shows a case where the sum [m1 + m2 + m3 + m4 + m5] of the signal strength mj is equal to or greater than the threshold value E1. FIG. 15B shows a case where the sum [m1 + m2 + m3 + m4 + m5] of the signal strength mj is less than the threshold value E1.

  Thereby, the signal processing system 2 can reduce misrecognition. For example, it is assumed that the object recognition unit 261 identifies a detection target or a detection target event based on a normalized frequency distribution of the normalized intensity nj. In this case, the object recognizing unit 261 has the normalized strength nj even if there is no detection target object or detection target event in the detection area of the sensor 1 and noise (dark noise or noise generated in the surroundings) is input. There is a possibility that the frequency distribution is misrecognized by determining that the frequency distribution is similar to the characteristics when there is a detection target object or a detection target event in the detection area. Therefore, the signal processing system 2 suppresses misrecognition by determining whether or not recognition processing is possible using the signal strength mj before normalization.

  Alternatively, all the target banks 5a may be divided into a plurality of filter bank groups 50 (see FIG. 16). In this case, the signal processing system 2 may determine whether or not the sum of the signal strengths mj before normalization is equal to or greater than the threshold value E2 in each of the plurality of filter bank groups 50. That is, the signal processing system 2 performs the recognition processing by the object recognition unit 261 only when the sum of the signal strengths mj before normalization is equal to or greater than the threshold E2 in any filter bank group 50, or the object recognition unit 261. The result of recognition processing by is made valid. Alternatively, the signal processing system 2 performs the recognition processing by the object recognition unit 261 only when the sum of the signal strengths mj before normalization is equal to or greater than the threshold E2 in all the filter bank groups 50, or by the object recognition unit 261. The result of the recognition process may be valid.

  The level setting unit 284 and the parameter adjustment unit 285 constitute a third erroneous recognition prevention processing unit (see FIG. 1). The parameter adjustment unit 285 changes a parameter for adjusting the detection sensitivity of the detection target object or the detection target event in the recognition processing of the object recognition unit 261. Parameters for adjusting the detection sensitivity include the above-described threshold values E1, E2, and the like. Hereinafter, when the threshold values E1 and E2 are not distinguished, they are referred to as a threshold value E0.

  Next, the signal processing system 2 includes a state machine for performing the above-described processes, and FIG. 17 shows a basic operation of the state machine (operation when the level setting unit 284 is not used).

  First, when the power is turned on or immediately after reset is released, the state machine starts operation from the idle state J11. Then, the state machine transits from the idle state J11 to the state I00 (t01).

  Here, the noise level in the surrounding environment where the sensor 1 is installed may change due to factors such as an increase or decrease in factors that change the noise level. Therefore, even if the threshold value E0 in the threshold value determination process is set once, if the noise level changes, the current setting value does not perform an expected operation, and erroneous recognition occurs or a detection object or a detection target event is detected. Despite being present, it may become undetected.

  Therefore, in the state I00 transitioned from the idle state J11, the parameter adjustment unit 285 performs an operation for setting the threshold value E0 in the threshold value determination process in the startup period, and after setting the threshold value E0, transitions to the state S11 (t02). . Specifically, in the state I00, the signal strength mj in each target bank 5a is measured. Then, the parameter adjustment unit 285 calculates a threshold value E0 by multiplying the average value of the signal strength mj of each target bank 5a by a predetermined coefficient, and uses this threshold value E0 as a threshold value in subsequent threshold value determination processing. Further, the range of the threshold value E0 is limited by a predetermined upper limit value and lower limit value. The upper limit value of the threshold value E0 is set to ensure the detection accuracy of the detection target object or the detection target event, and the lower limit value of the threshold value E0 is set to ensure the effect of suppressing erroneous recognition due to noise.

  Then, immediately after the start of the operation of the state machine, if it is considered that the detection object or the detection target event does not exist in the detection area of the sensor 1 and the sensor signal corresponding to the noise is output from the sensor 1, the state I00 The set threshold value E0 is a value based on noise.

  As described above, in the state machine of FIG. 17, the parameter adjustment unit 285 sets the threshold value E0 used for the threshold value determination process in accordance with the ambient noise environment during the startup period. Specifically, instead of performing recognition processing immediately after activation, first, the ambient noise level is measured from the sensor signal, and the parameter adjustment unit 285 multiplies the measured value by a predetermined coefficient to set the threshold value E0. Calculated. Therefore, even when the surrounding environment of the sensor 1 changes and the noise level also changes, the threshold value E0 can be appropriately changed during the startup period, so that erroneous recognition due to noise can be reduced.

  Then, the state machine transits from the state I00 to the state S11 (t02), and in the state S11, the object recognition unit 261 detects a detection target object or a detection target event (hereinafter referred to as a detection state). For example, the state transits to state W11 (t03). On the other hand, in the state S11, if the object recognition unit 261 does not detect the detection target object or the detection target event (hereinafter referred to as a non-detection state), a predetermined time after the transition to the state S11. After elapses, the state transitions to state S16 (t04). And if it is a non-detection state in state S16, it will change to state S11 (t05). That is, when the non-detection state continues from the state S11, a transition reciprocating between the state S11 and the state S16 is indicated.

  And if it will be in a detection state in state S11, S16, it will change to state W11 (t03, t06), and after waiting for a predetermined time in state W11, it will change to state S12 (t07), and will not change to state S13. Transitions under conditions (t08). In the state S13, if it is a non-detection state or if the detection state continues for a predetermined time or more, the state transitions to the state S14 (t09). And if it is a detection state in state S14, it will change to state S13 (t10). That is, when the detection state continues from state S13, a transition that reciprocates between state S13 and state S14 is indicated.

  And if it is a non-detection state in state S14, it will change to state S15 (t11). If it is a non-detection state in state S15, it will change to state S11 (t12), and if it is a detection state in state S15, it will change to state W11 (t13).

  That is, if it is a non-detection state, it will change the state S11 vicinity, and if it is a detection state, it will change the state S13 vicinity. Then, the signal processing system 2 performs the above-described processes while changing the states of the state machine.

  In the state machine shown in FIG. 17, when the detection state continues from state S13, a transition that reciprocates between state S13 and state S14 is shown. Therefore, when a transition that reciprocates between the state S13 and the state S14 occurs, the duration of the detection state can be determined by counting the number of times that the state S14 is passed. Therefore, if the upper limit of the duration of the detection state or the upper limit of the number of times of passing through the state S14 is set in advance and the upper limit is exceeded in the state S14, the state does not transit to the state S13 even in the detection state. Transition to the state I11 (t14).

  In the state I11, since the threshold value E0 used in the threshold value determination process is too small, the parameter adjustment unit 285 performs an operation of resetting the threshold value E2, assuming that the detection state may be continued. Specifically, in the state I11, the signal strength mj in each target bank 5a is measured. Then, the parameter adjustment unit 285 calculates the threshold value E0 by multiplying the average value of the signal strength mj of the target bank 5a by a predetermined coefficient, and uses this threshold value E0 in the subsequent threshold value determination processing. Further, the range of the threshold value E0 is limited by a predetermined upper limit value and lower limit value.

  Note that the parameter adjustment unit 285 resets the threshold E0 newly calculated in the state I11 to the threshold E0 newly calculated in the state I11 instead of the threshold E0 in use only when the threshold E0 newly calculated in the state I11 is larger than the threshold E0 currently in use. Set. Conversely, the parameter adjustment unit 285 does not reset the threshold E0 newly calculated in the state I11 to the new threshold E0 calculated in the state I11 when the threshold E0 newly calculated in the state I11 is equal to or less than the threshold E0 currently used. Continue to use E0. And after the process in the state I11 is completed, it changes to state S11 (t15).

  In the state machine shown in FIG. 17, when the non-detection state continues from the state S11, a transition that reciprocates between the state S11 and the state S16 is shown. Therefore, when a transition that reciprocates between the state S11 and the state S16 occurs, the duration of the non-detection state can be determined by counting the number of times that the state S16 is passed. Therefore, if the upper limit of the duration of the non-detection state or the upper limit of the number of times of passing through the state S16 is set in advance and the upper limit is exceeded in the state S11, the state transitions to the state S16 even in the non-detection state. There is no transition to state I12 (t16).

  In the state I12, the parameter adjustment unit 285 performs an operation of resetting the threshold value E0 because the non-detection state may be continued because the threshold value E0 used in the threshold value determination process is too large. Specifically, in the state I12, the signal strength mj in each filter bank 5a is measured. Then, the parameter adjustment unit 285 calculates the threshold value E0 by multiplying the average value of the signal strength mj of the target bank 5a by a predetermined coefficient, and uses this threshold value E0 in the subsequent threshold value determination processing. Further, the range of the threshold value E0 is limited by a predetermined upper limit value and lower limit value.

  Note that the parameter adjustment unit 285 resets the threshold E0 newly calculated in the state I12 to the threshold E0 newly calculated in the state I12 instead of the threshold E0 in use only when the threshold E0 newly calculated is smaller than the threshold E0 currently in use. Set. Conversely, when the threshold value E0 newly calculated in the state I12 is greater than or equal to the currently used threshold value E0, the parameter adjustment unit 285 does not reset the new threshold value E0 calculated in the state I12, and does not reset the threshold value in use. Continue to use E0. Then, after the processing in the state I12 is completed, the state transitions to the state S11 if it is a non-detection state (t17), and transitions to the state W11 if it is a detection state (t18).

  As described above, when the detection state or the non-detection state continues longer than the preset time, it is determined that the current threshold value E0 may be set to an inappropriate value for the current noise. The threshold value E0 is reset. Therefore, when erroneous recognition occurs due to the threshold E0 being too small, the detection sensitivity is lowered by updating to a larger threshold E0, and the occurrence of erroneous recognition can be suppressed. In addition, when the detection target or the detection target event cannot be detected because the threshold E0 is too large, the detection sensitivity is increased and detection omission can be reduced by updating to a smaller threshold E0.

  However, even if the threshold value E0 used in the threshold value determination process is updated as described above, erroneous recognition may occur due to a large change in the surrounding environment, or the detection target object or the detection target event may not be detected. There is a possibility that detection failure will occur.

  Therefore, the signal processing system 2 of this embodiment includes a level setting unit 284 (see FIG. 1). FIG. 18 shows the operation of the state machine using the level setting unit 284.

  The level setting unit 284 sets a sensitivity level indicating the level of detection sensitivity of the detection target object or the detection target event in the recognition processing of the object recognition unit 261. This level setting unit 284 sets the sensitivity level to a lower level when it is determined that the erroneous recognition by the object recognition unit 261 is likely to occur despite the update processing of the threshold value E0 in the states I11 and I12. To do. Furthermore, the level setting unit 284 sets the sensitivity level to a higher level when it is determined that erroneous recognition by the object recognition unit 261 is unlikely to occur.

  When the sensitivity level is high, the parameter adjustment unit 285 sets a parameter so that the detection sensitivity of the detection target object or the detection target event is high based on the sensitivity level set by the level setting unit 284, and the sensitivity level is low The parameter is set so that the detection sensitivity of the detection target object or the detection target event is low. That is, when the sensitivity level set by the level setting unit 284 is high, the parameter adjustment range (upper limit, lower limit) by the parameter adjustment unit 285 is set to a range in which the detection target object or the detection target event is relatively easy to detect. When the sensitivity level set by the level setting unit 284 is low, the parameter adjustment range (upper limit, lower limit) by the parameter adjustment unit 285 is set to a range in which the detection target object or the detection target event is relatively difficult to detect.

  FIG. 18 shows a state C11 provided between the state S11 and the state I12 in the state machine of FIG.

  And if it will be in a detection state in state S11, S15, S16, and I12, it will change to state W11 (t03, t06, t13, t18), will wait for predetermined time in state W11, and will change to state S12 ( t07).

  In the state S12, the level setting unit 284 performs sensitivity level update processing. Specifically, the level setting unit 284 determines whether the detection state that causes the transition to the state W11 has occurred because the detection target object or the detection target event has been detected, or the movement of an object other than the detection target object or the detection target event ( Determine whether it was caused by misrecognizing (disturbance). The determination processing of the level setting unit 284 is performed based on the recognition result of the object recognition unit 261 based on the current sensor signal. When the level setting unit 284 determines that the detection state that has caused the transition to the state W11 has occurred because the detection target object or the detection target event has been detected, it is difficult for the object recognition unit 261 to perform erroneous recognition. (Normal time). Further, when the level setting unit 284 determines that the detection state that causes the transition to the state W11 has occurred due to erroneous recognition of a motion (disturbance) other than the detection target object or the detection target event, the level setting unit 284 performs It is determined that the situation is likely to cause erroneous recognition (when a disturbance occurs).

  The level setting unit 284 has a function of setting a flag indicating the level of detection sensitivity of the detection target object or the detection target event in the recognition processing of the object recognition unit 261. Based on the result of the determination processing described above, Update flag settings. The level setting unit 284 sets the flag to “0” when it is determined that it is normal time, and sets the flag to “1” when it is determined that disturbance is occurring. The flag “0” corresponds to the sensitivity level: “high”, and the flag “1” corresponds to the sensitivity level: “low”.

  Note that the flag update process by the level setting unit 284 is preferably performed when the determination that the disturbance is occurring continues for a predetermined number of times or when the determination that the time is normal continues for a predetermined number of times. . Alternatively, in the flag update processing by the level setting unit 284, when the determination that the disturbance has occurred occurs more than a predetermined number of times within a predetermined period, or the determination that it is normal occurs more than a predetermined number of times within a predetermined period. It is preferable to carry out in such a case.

  That is, the level setting unit 284 sets the sensitivity level to a lower level when it is determined that erroneous recognition of the object recognition unit 261 is likely to occur due to disturbance. The level setting unit 284 sets the sensitivity level to a higher level (returns the sensitivity level to the original level) when it is determined that erroneous recognition by the object recognition unit 261 is unlikely to occur. When the sensitivity level setting process by the level setting unit 284 is completed, the state transitions from the state S12 to the state S13 (t08).

  Thereafter, based on the flag state (0 or 1), the parameter adjustment unit 285 sets the threshold value E0 so that the detection sensitivity is higher if the flag is “0” (setting at the normal time). If the flag is “1”, the parameter adjustment unit 285 sets the threshold E0 so that the detection sensitivity is lower (setting for occurrence of disturbance). That is, if the flag is “0”, the parameter adjustment unit 285 sets the adjustment range of the threshold E0 to a relatively low range, and if the flag is “1”, the parameter adjustment unit 285 sets the adjustment range of the threshold E0 to a relatively high range. To do.

  If the duration of the non-detection state exceeds the upper limit in state S11, the state transitions to state C11 without transitioning to state S16 even in the non-detection state (t16A). Also in the state C11, the level setting unit 284 updates the flag setting. Specifically, the level setting unit 284 determines whether the non-detection state that caused the transition to the state C11 occurs because the detection target object or the detection target event does not actually exist, or the detection target object or the detection target event is actually In spite of the above, it is determined whether or not the detection object or the detection target event has occurred because it was misrecognized. The determination processing of the level setting unit 284 is performed based on the recognition result of the object recognition unit 261 based on the current sensor signal. Then, when the level setting unit 284 determines that the non-detection state that has caused the transition to the state C11 has occurred because the detection target object or the detection target event does not actually exist, erroneous recognition by the object recognition unit 261 is performed. It is determined that the situation is unlikely to occur (normal time). In addition, the level setting unit 284 misrecognized that the detection state that caused the transition to the state W11 does not include the detection target object or the detection target event even though the detection target object or the detection target event actually exists. If it is determined that the error has occurred, it is determined that a situation in which erroneous recognition by the object recognition unit 261 is likely to occur (when a disturbance occurs).

  Then, the level setting unit 284 updates the flag setting based on the result of the determination process described above. The level setting unit 284 sets the flag to “0” when it is determined that it is normal time, and sets the flag to “1” when it is determined that disturbance is occurring. When the sensitivity level setting process by the level setting unit 284 is completed. Transition from the state C11 to the state I12 (t16B).

  Note that the flag update process by the level setting unit 284 is preferably performed when the determination that the disturbance is occurring continues for a predetermined number of times or when the determination that the time is normal continues for a predetermined number of times. . Alternatively, in the flag update processing by the level setting unit 284, when the determination that the disturbance has occurred occurs more than a predetermined number of times within a predetermined period, or the determination that it is normal occurs more than a predetermined number of times within a predetermined period. It is preferable to carry out in such a case.

  Thereafter, based on the flag state (0 or 1), the parameter adjustment unit 285 sets the threshold E0 so that the detection sensitivity is higher if the flag is “0”, and if the flag is “1”, The threshold E0 is set so that the detection sensitivity is lower. That is, if the flag is “0”, the parameter adjustment unit 285 sets the adjustment range of the threshold E0 to a relatively low range, and if the flag is “1”, the parameter adjustment unit 285 sets the adjustment range of the threshold E0 to a relatively high range. To do.

  Note that, when a disturbance is generated by the level setting unit 284, the normal determination process is a method based on pattern recognition using the distribution of the frequency components of the sensor signal, and based on past fluctuations in the sensor signal. There are methods for detecting the presence or absence of difficult features.

  Further, the sensor 1 may output a sensor signal including information on the distance to the detection target by changing the frequency of the transmission wave with the passage of time. In this case, the sensor 1 uses, for example, an FMCW (Frequency-Modulated Continuous-Wave) method. FIG. 19 shows an example of the operation of the sensor 1 that repeats the sweeping process of increasing and decreasing the frequency (transmission frequency) fs of the transmission wave. In the sweep process, the sweep frequency width Δfa and the sweep time T1 are determined.

If the distance between the sensor 1 and the object to be detected is L and the speed of light is C, the sensor 1 receives the reflected wave after Td = 2L / C. The frequency (reception frequency) fr of the reflected wave changes with the sweep frequency width Δfa and the sweep time T1, similarly to the transmission frequency fs. Then, the sensor 1 generates a beat signal having a frequency fb equal to the frequency difference between the transmission frequency fs and the reception frequency fr, and outputs it as a sensor signal. The frequency fb of the beat signal is fb = [(Δfa · 2L) / (C · T1)]. Therefore, the distance L to the human body 200 is expressed by equation (14).
L = (fb · C · T1) / (2 · Δfa) (14) Then, the object recognition unit 261 can obtain the distance L to the detection target based on the equation (14). Furthermore, the object recognizing unit 261 can determine the movement state of the detection target in the detection area based on the information on the distance L. Note that the light speed C, the sweep time T1, and the sweep frequency width Δfa are known, and the object recognition unit 261 stores in advance each data of the light speed C, the sweep time T1, and the sweep frequency width Δfa.

  In FIG. 19, the sensor 1 alternately repeats a sweep time T1 for transmitting radio waves and a pause time T2 for not transmitting radio waves. In this case, assuming that the cycle of the sweep process is the processing cycle T0, the processing cycle T0 is the sum of the sweep time T1 and the pause time T2. Further, the sensor 1 may repeat the sweep process every sweep time T1. In this case, the processing cycle T0 and the sweep time T1 are equal. Note that the processing cycle T0 is preferably set within a range of 10 (ms) to 100 (ms), for example.

  Further, the signal processing system 2 includes an area setting unit 291. The area setting unit 291 sets the size or range of a detection area that can identify a detection target or a detection target event. Note that the size and range of the detection area include the concepts of the shape and area of the detection area.

  For example, when the sensor 1 is a sensor using the FMCW method, the bank signals Y1-Y5 of each target bank 5a are represented by beat signals. That is, each frequency of the bank signals Y1-Y5 includes information on the distance L to the detection target. As a result, the area setting unit 291 can set the size or range of the detection area to a desired size or range by attenuating components in a specific frequency range from the bank signals Y1-Y5. Specifically, the area setting unit 291 may be configured with a low pass filter when setting a detection area where the distance L is equal to or less than a predetermined distance. Moreover, the area setting part 291 may be comprised with a high pass filter (High Pass Filter), when setting the detection area where the distance L becomes more than predetermined distance. Moreover, the area setting part 291 may be comprised with a band pass filter (Band Pass Filter), when setting the detection area where the distance L is in the predetermined range.

  In addition, the signal processing system 2 provides a dielectric lens on at least one of the transmission antenna and the reception antenna of the sensor 1 instead of the area setting unit 291 and sets the size or range of the detection area with the dielectric lens. Also good.

  For example, the transmission antenna and the reception antenna may have individual directivities, and a dielectric lens may be provided on at least one or both of the transmission surface of the transmission antenna and the reception surface of the reception antenna. In this case, a detection area is formed by the directivity of the transmission antenna and the reception antenna and the dielectric lens. That is, the detection area is formed as a result of the combination of the directivity of each antenna and the characteristics of the dielectric lens. At this time, if the dielectric lens is used for both the transmitting surface of the transmitting antenna and the receiving surface of the receiving antenna, the dielectric lens is a separate component for the transmitting surface of the transmitting antenna and the receiving surface of the receiving antenna. It may be an integral part.

  The sensor 1 may include a dielectric lens on the transmission surface of the transmission antenna. In this case, the transmitting antenna and the receiving antenna are antennas having specific directivities. The radio wave emitted from the transmission antenna is refracted by the dielectric lens, and the transmission intensity is changed depending on the transmission direction of the radio wave. That is, the detection area is formed by the dielectric lens.

  The sensor 1 may include a dielectric lens on the receiving surface of the receiving antenna. In this case, the transmitting antenna and the receiving antenna are antennas having specific directivities. The reflected wave is refracted by the dielectric lens, and the gain changes depending on the direction of reception of the reflected wave. That is, the detection area may be controlled such that the size or range of the detection area is set by a mechanical structure such as a dielectric lens.

  Furthermore, the signal processing system 2 includes a movement identification unit 292 (see FIG. 1). The movement identification unit 292 identifies the movement state or fluctuation state of the object.

  For example, when the sensor 1 is a sensor using the FMCW method, the movement identification unit 292 periodically obtains the distance L to the object based on the frequency fb of the beat signal (sensor signal). Then, the movement identification unit 292 can obtain the movement trajectory of the object as the movement state or the fluctuation state by arranging the distances L to the object obtained periodically in order of time series. The movement locus data includes information on the moving direction and moving speed of the object.

  Further, when the sensor 1 is a sensor using the FMCW method or a Doppler sensor, the movement identification unit 292 indicates the movement state or fluctuation state of the object based on the time change of the phase difference between the transmission wave and the reflection wave. It is also possible to identify.

  Each of FIG. 20A, FIG. 20B, and FIG. 20C represents the time change of the phase difference between the transmission wave and the reflected wave. Here, the presence / absence of an object in the detection area and the window setting value Ta (see FIG. 4), which is the value of the time length of the time window W described above, are used as operating conditions, and each of FIGS. 20A, 20B, and 20C The operating conditions are different from each other.

  In FIG. 20A, as an operation condition, an object in a fine movement state (vibration state) exists at a predetermined distance (for example, 4 m) from the sensor 1 in the detection area, and the window setting value Ta is set to 1 second, and the phase difference of The time change is represented as a phase difference waveform Y41.

  In FIG. 20B, as an operation condition, an object in a fine movement state in the detection area is present at a predetermined distance (for example, 4 m) from the sensor 1 and the window setting value Ta is set to 3 seconds. It is represented as a phase difference waveform Y42.

  In FIG. 20C, as an operating condition, no object is present in the detection area, and the window setting value Ta is set to 3 seconds, and the time change of the phase difference is represented as a phase difference waveform Y43.

  The phase difference waveforms Y41 and Y42 are waveforms when an object in a fine movement state exists in the detection area, and the phase difference waveform Y43 is a waveform when no object exists in the detection area. The phase difference between the phase difference waveforms Y41 and Y42 changes periodically, and the phase difference of the phase difference waveform Y43 becomes substantially constant. Therefore, the movement identifying unit 292 can identify the finely-moving object (the object fluctuation state) in the detection area based on the period of time variation of the phase difference. That is, the movement identifying unit 292 can identify whether or not the object in the detection area is in a fine movement state based on the period of time variation of the phase difference.

  Further, in the operating condition when the phase difference waveform Y42 is obtained, the window setting value Ta is set to the same value as the half cycle of the fine movement (vibration) of the object. Further, in the operating condition when the phase difference waveform Y41 is obtained, the window setting value Ta is set to a value that is significantly smaller than the half cycle of fine movement (vibration) of the object. Accordingly, the amount of change in the phase difference in the phase difference waveform Y42 is larger than that in the phase difference waveform Y41. That is, by setting the window setting value Ta according to the period of fine movement (vibration) of the object, the movement identification unit 292 can easily identify the movement state or fluctuation state of the object in the detection area.

  The sensor 1 may be a two-channel Doppler sensor. In this case, the sensor 1 generates an I (In-phase) signal and a Q (Quadrature-phase) signal having a phase difference of π / 2 by quadrature detection of the received wave, and the I signal and the Q signal are detected by the sensor. Output as a signal. If the object moves in a direction approaching the sensor 1, the I signal advances by π / 2 from the Q signal, and if the object moves in a direction away from the sensor 1, the I signal is delayed by π / 2 from the Q signal. Therefore, the movement identification unit 292 can identify the approach and separation of the object, and can determine the approach and separation of the object as the movement state.

  The above-described processes of the signal processing system 2 are represented by the flowcharts of FIGS.

  FIG. 21 shows an outline of the above-described series of processing by the signal processing system 2.

  First, the A / D conversion unit 22 performs an A / D conversion process of converting the sensor signal amplified by the amplification unit 21 into a digital sensor signal and outputting it (X1). Next, the frequency analysis unit 23 performs a frequency conversion process of converting the sensor signal output from the A / D conversion unit 22 into a frequency domain sensor signal (frequency axis signal) by, for example, DCT (X2). Next, the frequency analysis unit 23 performs a filter bank process for extracting a signal for each filter bank 5a from the sensor signal in the frequency domain (X3).

  Next, the signal strength processing unit 24 determines whether or not the sum of the signal strengths mj is greater than or equal to a threshold value E0 (X4). If the sum of the signal intensities mj is greater than or equal to the threshold value E0, the background signal removal processing by the background signal removal unit 283 is performed in order to perform the recognition processing by the object recognition unit 261 (X5). If the sum of the signal intensities mj is less than the threshold value E0, the recognition processing by the object recognition unit 261 is not performed, and thus the above-described series of processing by the signal processing system 2 ends.

  Next, the signal intensity processing unit 24 applies each value of the window setting value Ta and the slide time Ts set by the time setting unit 25 to parameters of the second smoothing process by the smoothing unit 241 (X6). Then, the signal strength processing unit 24 performs signal strength processing on the sensor signal (X7). In the signal strength process, a first smoothing process, a second smoothing process, and a normalization process are performed. First, the signal strength processing unit 24 obtains the signal strength mji for each target bank 5a by executing a first smoothing process on the sensor signal. Then, the signal strength processing unit 24 performs the second smoothing process and the normalization process on the signal strength mji for each target bank 5a, thereby obtaining the standardized strength nj for each target bank 5a. That is, the signal strength processing unit 24 normalizes the signal strength for each target bank 5a and obtains the normalized strength nj.

  Then, the object recognizing unit 261 extracts the characteristics of the frequency distribution of the sensor signal determined from the normalized strength nj for each target bank 5a, or the characteristics of the component ratio of the sensor signal determined from the normalized strength nj for each target bank 5a. The object recognizing unit 261 performs a recognition process based on the extracted feature to determine whether or not it can be regarded as a detection target object or a detection target event (X8). And the output part 27 performs the output process of a detection signal according to the result of a recognition process (X9).

  FIG. 22 shows the details of the second smoothing process using the time window W in the signal intensity process (X7) described above.

  The smoothing unit 241 sets the value of the time length of the time window W to the window setting value Ta by the time setting unit 25 (X11) (see FIG. 4). Then, the smoothing unit 241 applies the time window W to the bank signal Yj for each target bank 5a, and extracts the signal intensities mji of a plurality of samples that fall within the time window W (X12). The smoothing unit 241 performs the above-described second smoothing process using the extracted signal strength mji (X13). Each time the smoothing unit 241 executes the second smoothing process, the smoothing unit 241 shifts the time window W applied to the bank signal Yj by the predetermined slide time Ts in the direction of time advance (X14). Then, the smoothing unit 241 again extracts signal intensities of a plurality of samples that fall within the time window W (X11). Thereafter, the smoothing unit 241 performs the second smoothing process using the time window W by repeating the processes of Steps X11 to X14.

  As described above, the signal processing system 2 of the first aspect according to the embodiment includes the frequency analysis unit 23, the signal intensity processing unit 24, the recognition unit 26, and the time setting unit 25. The frequency analysis unit 23 receives the sensor signal from the sensor 1 that receives the radio signal reflected by the object, and obtains the intensity of the sensor signal in each of the plurality of filter banks 5a as the signal intensity mji. The signal strength processing unit 24 sets at least two filter banks 5a among the plurality of filter banks 5a as target banks 5a. The signal intensity processing unit 24 performs a smoothing process (second smoothing process) for smoothing the signal intensity mji within the time window W applied to the bank signal Yj representing the time change of the signal intensity mji for each target bank 5a, and the signal A normalization process for normalizing the intensity mji is executed to obtain the normalized intensity nj of the sensor signal for each target bank 5a. The recognizing unit 26 detects the object (detection target) or the movement of the object (detection target event) based on at least one of the frequency distribution of the sensor signal determined from the standardized intensity nj for each target bank 5a and the component ratio of the standardized intensity nj. A recognition process for identifying is performed. The time setting unit 25 can variably set the time length of the time window W (window setting value Ta). Then, the signal intensity processing unit 24 executes a smoothing process every time the time window W is shifted by a predetermined slide time Ts with respect to the bank signal Yj.

  In the signal processing system 2 described above, the second smoothing process is executed so that the detection target object or the movement of the detection target object can be accurately identified by the setting process of the window setting value Ta by the time setting unit 25. The identification accuracy of the recognition unit 26 is improved. In addition, the smoothing unit 241 executes the second smoothing process every time the time window is shifted in the time progress direction by the slide time Ts, so that the recognition unit 26 can be identified by appropriately setting the slide time Ts. The accuracy can be further improved. Therefore, the signal processing system 2 can maintain the identification accuracy of the object or the movement of the object regardless of the size, shape, or movement of the object.

  Moreover, the signal processing system 2 of the 2nd aspect which concerns on embodiment is further provided with the misrecognition prevention process part 28 for reducing the misrecognition in which the recognition part 26 recognizes an object accidentally in the 1st aspect. Is preferred.

  The signal processing system 2 described above can reduce erroneous recognition of the recognition unit 26.

  Moreover, it is preferable that the signal processing system 2 of the 3rd aspect which concerns on embodiment is further provided with the signal component extraction part 281 and the disturbance judgment part 282 in a 2nd aspect. The signal component extraction unit 281 obtains a signal component resulting from the movement of the object from the normalized strength nj for each target bank 5a. The disturbance determination unit 282 recognizes by the recognition unit 26 when the magnitude of fluctuation of the signal component per unit time is not less than a predetermined value or not more than a predetermined value in at least one target bank 5a out of at least two target banks 5a. Prohibit processing. Alternatively, the disturbance determination unit 282 invalidates the result of the recognition process performed by the recognition unit 26.

  The signal processing system 2 described above can accurately detect the movement of the detection target object or the detection target object regardless of the presence or absence of a sharp temporal change in the signal caused by the movement of the object other than the detection target. Furthermore, the signal processing system 2 can be simplified and downsized.

  Moreover, it is preferable that the signal processing system 2 of the 4th aspect which concerns on embodiment is further provided with the background signal removal part 283 in a 2nd or 3rd aspect. When recognizing the object or the movement of the object, the recognizing unit 26 at least one of the characteristics of the frequency distribution of the sensor signal determined from the normalized intensity nj for each target bank 5a and the characteristics of temporal continuity of the frequency distribution. To further identify stationary background signals. The background signal removal unit 283 removes the background signal from the signal for each target bank 5a.

  The signal processing system 2 described above can reduce misrecognition caused by the movement of an object other than the detection target.

  The signal processing system 2 according to the fifth aspect according to the embodiment preferably further includes a level setting unit 284 and a parameter adjustment unit 285 in any one of the second to fourth aspects. The level setting unit 284 sets a sensitivity level indicating the level of detection sensitivity with respect to an identification target of recognition processing. The parameter adjustment unit 285 changes a parameter for adjusting the detection sensitivity. Then, the parameter adjustment unit 285 sets the parameter so that the detection sensitivity is high when the sensitivity level is high, and sets the parameter so that the detection sensitivity is low when the sensitivity level is low.

  The signal processing system 2 described above can reduce misrecognition caused by the movement of an object other than the detection target while balancing the improvement in detection sensitivity and the reduction in misrecognition.

  In the signal processing system 2 of the sixth aspect according to the embodiment, in any one of the first to fifth aspects, the recognizing unit 26 recognizes each target bank 5a included in a predetermined period Tr. It is preferable to identify the object or the movement of the object based on the characteristics of the normalized intensity nj.

  The signal processing system 2 described above can accurately identify the object to be detected or the movement of the object to be detected.

  Further, in the signal processing system 2 of the seventh aspect according to the embodiment, in the sixth aspect, when the recognition unit 26 performs the recognition processing based on the frequency distribution, the time change of the normalized strength nj in the target bank 5a Whether or not a specific frequency component is included in the signal representing the value is preferably a characteristic of the normalized strength nj.

  The signal processing system 2 described above can accurately identify the detection target object or the movement of the detection target object with a simple process.

  Further, in the signal processing system 2 of the eighth aspect according to the embodiment, in the sixth aspect, when the recognition unit 26 performs the recognition process based on the frequency distribution, the recognition unit 26 for each target bank 5a included in the predetermined period Tr. Of the normalized strengths nj, the number of times that at least one normalized strength nj has changed across at least one threshold value R1, R2 is preferably a characteristic of the normalized strength.

  The signal processing system 2 described above can accurately identify the detection target object or the movement of the detection target object with a simple process.

  The signal processing system 2 of the ninth aspect according to the embodiment is the size or range of the detection area that can identify the object or the movement of the object in any one of the first to eighth aspects. It is preferable to further include an area setting unit 291 for setting.

  Since the signal processing system 2 described above can set a detection area according to the detection target object or the movement of the detection target object, the application range of the signal processing system 2 is expanded.

  The signal processing system 2 according to the tenth aspect according to the embodiment further includes a movement identification unit 292 that identifies the movement state of the object or the fluctuation state of the object in any one of the first to ninth aspects. It is preferable. The sensor 1 transmits a radio signal as a transmission wave, and receives a radio signal reflected by an object as a reflected wave. Then, the movement identification unit 292 uses the time change of the phase difference between the transmitted wave and the reflected wave caused by the movement or fluctuation of the object, or the time change of the distance to the object obtained based on the sensor signal. Identifies a state or variation state.

  The signal processing system 2 described above can obtain a recognition result corresponding to the moving state of the object by further identifying the moving state of the object.

  In the signal processing system 2 of the eleventh aspect according to the embodiment, in any one of the first to tenth aspects, the slide time Ts is based on the time length of the time window W (window setting value Ta). Small is preferable.

  In the signal processing system 2 described above, a part of the time window W1 before shifting in the time progressing direction overlaps with the time window W2 after shifting in the time progressing direction in the time axis direction. Therefore, the feature of the sensor signal is extracted more reliably.

  The program of the twelfth aspect according to the embodiment causes the computer system to function as the frequency analysis unit 23, the signal strength processing unit 24, the recognition unit 26, and the time setting unit 25. The frequency analysis unit 23 receives the sensor signal from the sensor 1 that receives the radio signal reflected by the object, and obtains the intensity of the sensor signal in each of the plurality of filter banks 5a as the signal intensity mji. The signal strength processing unit 24 sets at least two filter banks 5a among the plurality of filter banks 5a as target banks 5a. The signal intensity processing unit 24 performs a smoothing process (second smoothing process) for smoothing the signal intensity mji within the time window W applied to the bank signal Yj representing the time change of the signal intensity mji for each target bank 5a, and the signal A normalization process for normalizing the intensity mji is executed to obtain the normalized intensity nj of the sensor signal for each target bank 5a. The recognizing unit 26 detects the object (detection target) or the movement of the object (detection target event) based on at least one of the frequency distribution of the sensor signal determined from the standardized intensity nj for each target bank 5a and the component ratio of the standardized intensity nj. A recognition process for identifying is performed. The time setting unit 25 can variably set the time length of the time window W (window setting value Ta). Then, the signal intensity processing unit 24 executes a smoothing process every time the time window W is shifted by a predetermined slide time Ts with respect to the bank signal Yj.

  The above-described program can cause a computer system to execute each function of the above-described signal processing system 2. Therefore, the program can maintain the identification accuracy of the object or the movement of the object regardless of the size, shape, movement or the like of the object.

  The above-described embodiment is an example of the present invention. For this reason, the present invention is not limited to the above-described embodiment, and various modifications can be made depending on the design and the like as long as the technical idea according to the present invention is not deviated from this embodiment. Of course, it is possible to change.

DESCRIPTION OF SYMBOLS 1 Sensor 2 Signal processing system 23 Frequency analysis part 24 Signal strength processing part 25 Time setting part 26 Recognition part 261 Object recognition part 262 Background recognition part 28 False recognition prevention processing part 281 Signal component extraction part 282 Disturbance judgment part 283 Background signal removal part 284 Level setting unit 285 Parameter adjustment unit 291 Area setting unit 292 Movement identification unit 5a Filter bank, target bank mji Signal strength nj Normalization strength Yj Bank signal W Time window Ta Window setting value Ts Slide time Tr Predetermined period R1, R2 Threshold

Claims (12)

A frequency analysis unit that receives a sensor signal from a sensor that receives a radio signal reflected by an object, and obtains the intensity of the sensor signal in each of a plurality of filter banks as a signal intensity;
A smoothing process for smoothing the signal strength within a time window applied to a bank signal representing a time change of the signal strength for each target bank, wherein at least two filter banks of the plurality of filter banks are each set as a target bank; And a signal strength processing unit for performing a standardization process for normalizing the signal strength, and obtaining a standardized strength of the sensor signal for each target bank,
A recognition unit for performing recognition processing for identifying the object or the movement of the object by at least one of a frequency distribution of the sensor signal determined from the normalized intensity for each target bank, or a component ratio of the normalized intensity;
A time setting unit capable of variably setting the time length of the time window,
The signal strength processing unit executes the smoothing process every time the time window is shifted by a predetermined slide time with respect to the bank signal. The signal processing system according to claim 1, further comprising a misrecognition prevention processing unit for reducing misrecognition by which the recognition unit erroneously recognizes the object. The erroneous recognition prevention processing unit includes a signal component extraction unit and a disturbance determination unit,
The signal component extraction unit obtains a signal component resulting from the movement of the object from the normalized intensity for each target bank,
When the magnitude of the fluctuation per unit time of the signal component is equal to or greater than a predetermined value or less than a predetermined value in at least one target bank of the at least two target banks, the disturbance determination unit The signal processing system according to claim 2, wherein a recognition process is prohibited or a result of the recognition process by the recognition unit is invalidated. The erroneous recognition prevention processing unit has a background signal removal unit,
When the recognition unit identifies the object or the movement of the object, the frequency distribution characteristic of the sensor signal determined from the normalized intensity for each target bank, and the temporal continuity characteristic of the frequency distribution, Further identifying a stationary background signal by at least one of
The signal processing system according to claim 2, wherein the background signal removing unit removes the background signal from the signal for each target bank. The erroneous recognition prevention processing unit includes a level setting unit and a parameter adjustment unit,
The level setting unit sets a sensitivity level indicating a level of detection sensitivity with respect to an identification target of the recognition process,
The parameter adjustment unit is configured to change a parameter for adjusting the detection sensitivity, and when the sensitivity level is high, the parameter is set so that the detection sensitivity is high, and when the sensitivity level is low, The signal processing system according to any one of claims 2 to 4, wherein the parameter is set so that the detection sensitivity is low. The said recognition part identifies the said object or the motion of the said object based on the characteristic of the said normalization intensity | strength for every said object bank contained in the predetermined period decided beforehand. A signal processing system according to claim 1. When the recognition unit performs the recognition process based on the frequency distribution, whether the signal representing the temporal change in the normalized strength in the target bank includes a specific frequency component is determined based on the normalized strength. The signal processing system according to claim 6, wherein the signal processing system is characterized. When the recognition unit performs the recognition process based on the frequency distribution, at least one standardized strength of the target banks included in the predetermined period crosses at least one threshold. The signal processing system according to claim 6, wherein the number of times of change is a characteristic of the normalized strength. The signal processing system according to claim 1, further comprising an area setting unit that sets a size or a range of a detection area that can identify the object or a motion of the object. . A movement identification unit for identifying a movement state of the object or a variation state of the object;
The sensor transmits the radio signal as a transmission wave, receives the radio signal reflected by the object as a reflected wave,
The movement identification unit uses a time change of a phase difference between the transmitted wave and the reflected wave caused by movement or fluctuation of the object, or a time change of a distance to the object obtained based on the sensor signal. The signal processing system according to claim 1, wherein the movement state or the fluctuation state is identified. The signal processing system according to any one of claims 1 to 10, wherein the slide time is smaller than the time length of the time window. Computer system
A frequency analysis unit that receives a sensor signal from a sensor that receives a radio signal reflected by an object, and obtains the intensity of the sensor signal in each of a plurality of filter banks as a signal intensity;
A smoothing process for smoothing the signal intensity within a time window applied to a bank signal representing a time change of the signal intensity for each target bank, wherein at least two filter banks of the plurality of filter banks are each set as a target bank; And a signal strength processing unit that performs a standardization process for normalizing the signal strength, and obtains a standardized strength of the sensor signal for each target bank,
A recognition unit for performing recognition processing for identifying the object or the movement of the object by at least one of a frequency distribution of the sensor signal determined from the normalized intensity for each target bank, or a component ratio of the normalized intensity;
Function as a time setting unit capable of variably setting the time length of the time window;
The signal strength processing unit executes the smoothing process every time the time window is shifted by a predetermined slide time with respect to the bank signal.