JP2001272458A - Data time-series merging method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data time-series merging method which is higher in the accuracy of merging than before and reduces the load of operation on an operator by automation. SOLUTION: According to the output data from a sound source which have been passed through a signal processor 1 including a sound sensor, a signal detecting and tracking device 2, a data integrating device 3, a target discriminating device 4, and a target movement analyzing device 5, a data merging decision device 6 merges two pieces of sound source information succeeding in the undetactable period of sound source information temporally on condition that two entity names found from the sound source information in the observation periods before and after the undetectable period of the sound source information match each other and are both larger in accuracy than a prescribed value and the azimuth component of the sound source information detected in the observation period after the undetectable period is included in the detection predicted azimuth range of the sound source which is found in the observation period of the undetectable period and a specific time later.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process for automatically detecting and tracking a plurality of signals obtained from a single sound source using an acoustic sensor. If the sound source information is once detected and the sound source information is detected again within a certain period thereafter, it is automatically determined whether or not the two sound source information before and after the sound source information undetectable period have the same target. The present invention relates to a data time series combining method for temporally combining two sound source information based on a result of the determination.

[0002]

2. Description of the Related Art In a process of automatically detecting and tracking a plurality of signals obtained from one sound source by using an acoustic sensor, data of a processing result which is originally the same sound source is set to a plurality of same targets by an observation time. Is recognized as
Conventionally, the operator of the sensor system manually specifies multiple targets mainly based on the result of the automatic detection and tracking process,
Time-series joining of data was performed.

FIG. 5 is an explanatory view of a conventional operation for joining detection targets. In the automatic detection and tracking processing, information relating to the frequency of the sound source and information relating to the azimuth can be obtained. However, the operator performs the combining operation while particularly viewing the information relating to the azimuth. FIG.
The operator recognizes that the two different targets A and C are the same target based on the azimuth information arranged in chronological order, and performs an operation of combining data. Note that the thick arrow in the figure frame reads “the left side of the arrow means the right side” or “if the left side of the arrow is the right side of the arrow”.

[0004]

However, the conventional target combining method as described above has the following problems (1) and (2). (1) Problem of Improving the Accuracy of Combining When There Are Multiple Target Candidates to be Combined FIG. 6 is an explanatory diagram of a problem in the conventional target combining. In the conventional method, as shown in FIG. 6, when there are a plurality of targets that can be combined only with the azimuth information, the operator who originally couples the target A and the target C is erroneously recognized as the target A due to erroneous recognition and erroneous operation. The target D may be combined. Therefore, in addition to the joining operation based on the azimuth information, which was conventionally performed by the operator, by adding data processing based on the frequency information obtained by the automatic detection and tracking process, a more plausible combination as a combination from multiple existing joining candidates Processing to narrow down was required.

(2) Problem of Automation of Operation There is a problem that the burden on the operator should be reduced by automatically performing the operation conventionally performed by the operator on a computer. Conventionally, in performing a data combining operation, frequency information of a target automatically detected and tracked has been provided to an operator.
However, the operator can refer to the information by changing the viewpoint by switching screens, performing an operation when calling the screen, or increasing the amount of information given to the operator due to the increased complexity of data. Load tends to increase. Therefore, in order to reduce these loads on the operator, it has been necessary to automate the operation.

[0006]

According to the data time series combining method of the present invention, during the observation period, the frequency of an acoustic signal from a sound source received via an acoustic sensor having directivity is analyzed and each received signal is analyzed. Obtain signal strength data for each direction and frequency, perform detection and tracking of sound source information based on the data, integrate direction information and frequency information determined to be the same sound source, and based on the integrated frequency information, the entity name of the sound source and In addition to obtaining the certainty factor, the target motion analysis based on the azimuth and frequency information obtained in time series is used to obtain the detection prediction azimuth range and the prediction accuracy of the sound source after a predetermined time, and the detection has been performed during the observation period. Sound source information once becomes undetectable,
If there is sound source information that can be detected again within a certain period thereafter, whether or not the two sound source information detected respectively in the observation periods before and after the undetectable period of the sound source information can be temporally combined is determined. (1) The two entity names obtained from the sound source information match during the observation period before and after the sound source information cannot be detected, and the certainty factor for each entity name is both a specified value or more. . (2) The azimuth component of the sound source information detected in the observation period after the undetectable period is included in the detection azimuth range after a predetermined time of the sound source obtained in the observation period before the undetectable period of the sound source information. . When both of the above-mentioned determination conditions (1) and (2) are satisfied, the two pieces of sound source information detected respectively in the observation periods before and after the non-detection period of the sound source information are temporally combined.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a configuration diagram of a data time series combining device according to Embodiment 1 of the present invention. In FIG. 1, reference numeral 1 denotes a signal processing device having a built-in acoustic sensor, inputting an acoustic signal from a sound source for each beam of the acoustic sensor, and obtaining a signal intensity distribution by Fourier transform or the like. Reference numeral 2 denotes a signal detection and tracking device. The signal detection and tracking device 2 is described in, for example, Japanese Patent Application Laid-Open No. 9-15032,
This is a device that performs peak detection and event detection processing using a neural network described in "Signal Frequency Tracking Method", and performs signal frequency tracking and azimuth tracking. 3 is a data integration device. The data integration device 3 converts a signal from a plurality of detected sound sources into a signal detection and tracking device 2 based on a data integration method described in, for example, Japanese Patent Application Laid-Open No. 9-49871, a data integration method and its device. This is a device that determines whether the signals are from the same sound source based on the Mahalanobis distance based on the azimuth information which is the result of the above, and integrates the azimuth information and the frequency information when it is determined that they are the same.

Reference numeral 4 denotes a target class identification device. The target class identification device 4 calculates a target target certainty factor of a sound source by comparing with a mechanism-specific catalog described in, for example, Japanese Patent Application Laid-Open No. 9-329661, a target automatic classification / identification method. (Representing the sound source as a certainty factor). Reference numeral 5 denotes a target motion analyzer. As described in, for example, “Japanese Unexamined Patent Application Publication No. 9-90012, Target Motion Analysis Method”, the target motion analysis device 5 generates a sound source (movement) based on frequency information and azimuth information that are the processing results of the data integration device 3. This is a device that performs a target motion analysis process in which the internal state quantities relating to the position and velocity of the body are estimated by a nonlinear least squares method or a Kalman filter. Reference numeral 6 denotes a data combination determination device according to the present invention. The data combination determination device 6 holds the classification result of the target classification device 4, and even if the target sound source cannot be detected once in the target motion analysis device 5, the motion analysis result up to that time is retained. If it has been obtained, the existence range when the sound source is detected again is predicted from the result. An apparatus that determines whether data can be combined from re-detected sound sources based on the classification result and the predicted existence range.

FIG. 2 is an operation explanatory diagram in which input and output signals are entered in the device configuration diagram of FIG. 1. First, an outline of the operation from the signal processing device 1 to the target motion analysis device 5 will be described with reference to FIG. I do. The signal processing device 1 has a built-in acoustic sensor (for example, one in which a plurality of receiving transducers are arranged in an array), and forms a directional receiving beam using the acoustic sensor. When the formation direction of the received beam is scanned to monitor a continuous azimuth range, a received signal from a sound source is obtained by the received beam group. The signal processing device 1 performs frequency analysis such as Fourier transform on a signal received by the received beam group, and obtains signal intensity data for each received beam (that is, each direction) and each frequency. The signal strength data for each direction and frequency is supplied to the signal detection and tracking device 2.

The signal detection and tracking device 2 detects a signal of a sound source using the input signal strength data and individually tracks the signals as frequency and azimuth data. The data of the detected frequency and azimuth signals is represented by (f ₁ , θ
₁ ), (f ₂ , θ ₂ ),..., (F _n , θ _n ).
The data integration device 3 integrates data determined to be signals from the same sound source from the data in the frequency and azimuth directions by an integration determination process regarding the Mahalanobis distance. Here, the data of the integration result at the time t is ((f ₁ , f
₂ ,..., F _m ), θ) _t . Here, θ is the azimuth component (θ ₁ ,
θ ₂ ,..., θ _n ). The data integration device 3 manages integration result data from the same sound source that is continuous in time in an integration result database. The data is represented by the following equation (1). Here, t ₀ is the time when the sound source is automatically detected _first , and t ₁ is the latest time when the sound source is automatically detected.

[0011]

(Equation 1)

In the target class identifying device 4, the data integrating device 3
(F ₁ , f ₂ ,..., F _m )
Based on Dempster-Scheffer's theory, what is the entity of the sound source (what is the entity having the frequency component as the sound source) is estimated. The entity name of the result is _Obj , and the likelihood (confidence) that was the basis of the estimation is CL
And

On the other hand, in the target motion analysis device 5, the data stored in the time-series data integration result database is used to determine T seconds (for example, 60 seconds described later) of the sound source.
The later predicted detection direction and its prediction accuracy are calculated. This target motion analysis processing is performed at regular processing intervals T. Doing motion analysis processing from the information A obtained in now time t _1,
The new information B obtained at the time t ₂ after T seconds is added to the above A, and thereby the target motion analysis is performed again. here,
T seconds in the predicted detection direction after T seconds of the sound source and the processing cycle T are the same time, and for example, 1 minute (60 seconds) is used.

The time-series data Σ ((f ₁ , f ₂ ,...,
f _m), theta) of the frequency components of _t, focusing on the typical first frequency, by performing data processing by the nonlinear least square method, it is possible to obtain the following time-series data. Σ (θ, Δθ, Δd / d, 1 / d, F) _t where θ: sound detection direction (Σ ((f ₁ , f ₂ ,..., F _m ),
θ) Equivalent to θ of _t ) Δθ: Rate of change in direction of sound source Δd: Rate of change in distance between sound source and acoustic sensor d: Distance between sound source and acoustic sensor F: Detection frequency group of sound source (f ₁ , f ₂ ,...) , F _m ). The following frequency (a) or (b) is used as a method of selecting the representative frequency. (A) Among the frequency groups, the frequency detected continuously for the longest time. (B) Among the frequency groups, the frequency with the highest signal strength. From this time-series data, T seconds of the sound source (for example,
Prediction and detection orientation theta _t the prediction accuracy Shigumashita _t after sec) can be calculated. Here, the prediction accuracy σθ _t, is a value in consideration of the observation error of the orientation due to noise observation area.

Next, the operation of the data combination judging device 6 will be described. As described above, the data combination determination device 6 includes:
The entity name O _{bj of the} sound source, which is the result of data processing from the signal processing device 1 to the target class identification device 4 and the target motion analysis device 5
And confidence associated therewith, detection prediction orientation theta _t the prediction accuracy Shigumashita _t of the sound source are supplied. The sound source at this time is set as a target A. It is assumed that the signal from the sound source of the target A temporarily stops at time t ₁ in this state, and the signal from the sound source is detected again at a certain time t ₂ within a certain period (for example, within 30 minutes described later). The data integration device 6 from the time t ₂
It is assumed that the result shown in FIG. 3 is the target B, and FIG. That is, FIG. 3 is a diagram showing azimuth information of two targets (targets A and B) arranged in time series. Here, the fixed period will be described. The target class identification device 4 has a memory for storing two-dimensional data of target direction × time as shown in FIG. 3, but the maximum time on the time axis is determined from this memory capacity. is there. Therefore 30
Since the memory is the data at the time of the minute elapsed is erased, (the upper limit of undetectable time of the sound source information) maximum allowable time from time t ₁ in FIG. 3 to t ₂ will be in this case 30 minutes.

Target A is entity O_bjSame as the process estimated
Also, for the target B, the data is integrated from the signal processing device 1.
Through the process of processing up to the device 3, the data of the integration result
((F '₁, F '_Two , ……, f '_l ), Θ ') and time series
Data Σ ((f '₁ , F '_Two , ……, f '_l ), Θ ')_t But
can get. Here, with respect to the target B, the target class identification device 4
By processing, the sound source entity O _{bj '}, The confidence CL '
can get. In this state, the data combination determination device 6
Information has been obtained. Entity name until information on target A is discontinued: O_bj Goal A is O_bjIs certainty: CL Detection predicted azimuth information of target A: θ_t Prediction accuracy of target A: σθ_t Entity name of target B after re-detection: O_{bj '} Goal B is O_bj'Confidence: CL' Azimuth component of integration result of target B: θ '

The data combination determination device 6 determines whether or not the following determination conditions (1) and (2) are both satisfied for the above information. (1) The two entity names _Obj and _Obj 'obtained from the respective sound source information match during the observation period before and after the period during which the sound source information cannot be detected, and both the certainty factors CL and CL' for each entity name are defined. Must be greater than or equal to the value. (2) The azimuth component of the detected sound source information and the observation period after the undetectable period are included in the detection predicted azimuth range after a predetermined time of the sound source obtained in the observation period before the undetectable period of the sound source information. . That is, it is determined whether or not the following equations (2) and (3) hold. _Obj = _Obj 'and CL, CL'> X (2) [theta] _{t- [} alpha] <[theta] <[theta] _t + [alpha] (3) where X is effective as a classification result determined by the characteristics of the system. Lower limit of certainty factor α: Error range of detected predicted orientation. When the above equations (2) and (3) are satisfied, the data combination determination device 6 combines the integration result data of the target A and the integration result data of the target B as in the following equation (4).

[0018]

(Equation 2)

As described above, according to the first embodiment, the following effects can be obtained by using the data combination determination device 6. (1) Improving the accuracy of the combination When combining the targets, a judgment is made as to whether or not the entities of the class identification result are the same as compared with those based on only the azimuth information that the operator has performed (determined) so far. In this way, it is possible to improve the accuracy (suppress the erroneous connection). (2) Reduction of operator operations Conventional operators themselves can recognize targets on a display device such as a display, and the steps that have been performed as integrated operations can be omitted by automatically performing operations on a computer, reducing operator operations. You can do it. (3) target motion analysis before binding of improving data in the data processing performance of the device did not perform only processing by the data up to time t ₀ ~t ₁ in FIG. 3, performs data combination based the data binding determination device As a result, the time t _{0 to} t
Data for a long time up to ₃ can be used for the calculation, and errors in the calculation result are reduced, and as a result, data processing performance is improved.

In the device configuration shown in FIGS. 1 and 2, an example in which a data combination determination device 6 is applied to a signal processing device 1, a signal detection and tracking device 2, a data integration device 3, a target class identification device 4, and a target motion analysis device 5 is shown. As shown, since the information from the target motion analysis device 5 was the predicted orientation of the target that was once lost, the data from the data integration device 3 was used instead of the target motion analysis device 5, and the target Data combination determination device 6 even if a device that predicts the appearance direction after T seconds is used.
The same effect can be obtained without changing. Further, in the present description, the target motion analysis device 5 has been described as a device that performs a calculation based on an azimuth and a frequency.
No. 12, Japanese Patent Publication No. 12-207, a target motion analysis device using only azimuth information is also proposed, and these devices may be used.

Embodiment 2 FIG. 4 is a configuration diagram of a data time series combining device according to Embodiment 2 of the present invention. FIG. 4 shows a configuration in which a spectrum correlation determination device 4A is used instead of the target class identification device 4 in FIG. 1, and the other devices are the same as those in FIG. The spectrum correlation determination device 4A of FIG. 4 focuses on the frequency component in the integration result of the data integration device 3, takes the correlation between the input spectra from the spectrum information, and determines whether or not the sound sources are the same. It is.

The operation of the apparatus shown in FIG. 4 will be described. The processing from the signal processing device 1 to the data integration device 3 in FIG. 4 is the same as in the first embodiment. Therefore, here, only the operation of the spectrum correlation determination device 4A and the operation of the data combination determination device 6 changed accordingly will be described.

[0023] In spectral correlation determination device 4A, frequency and orientation information which is input information from the data integration unit _{3 ((f 1, f 2} , ......, f m), θ) frequency component of the _t ((f _1, f _2, ......, attention is paid to the f _{m) t.} at a certain time t _0, the frequency information F1 (f1 ₁ input,
f1 _2, ......, f1 _i) and _t0, frequency information at a different time _{t 1 F2 (f2 1, f2} 2, ......, relative to f2 _{j) t1,} the following equation (5), (6) It is determined whether there is a subset of UF1 and UF2 such that UF1 = UF2 ... (5) In other _{words, UF1 = (f1 1, f1} 2, ......, fl k) UF2 = (f2 1, f2 2, ......, f2 l) k = l (5) the configuration of UF1 Indicates that the element and the component of UF2 match. _{R UF1, R UF2> R ...} (6) wherein R _UF1: ratio R _UF2 occupied subset UF1 of F1: Percentage is a subset UF2 of F2 R: a subset of setting by the system are valid This is the lower limit value to be determined. When the above equations (5) and (6) hold, the spectrum correlation determination device 4A assumes that there is a spectrum correlation between the two pieces of frequency information F1 and F2.

As described above, the combination information of the target and the time at which the spectrum correlation is recognized can be obtained. Here, unlike the case of the first embodiment, the data combination determination device 6 inputs a combination of a target having a high degree of spectral correlation and a time from the spectrum correlation determination device 4A,
The following information has been obtained: Time range in which target A was detected: T _{0 to} T ₁ Time range in which target B was detected: T _{2 to} T ₃ Time target combination in which correlation is established: ((A, T _i ),
(B, T _j )) Target predicted azimuth information of target A: θ _t Direction component of integration result of target B: θ ′ The data combination determination device 6 determines the above information with the following determination conditions (1) and (2). It is determined whether or not both are satisfied. (1) A correlation of a frequency spectrum can be obtained between two integrated frequency information obtained from the sound source information during an observation period before and after the sound source information cannot be detected. (2) The azimuth component of the sound source information detected in the observation period after the undetectable period is included in the detection azimuth range after a predetermined time of the sound source obtained in the observation period before the undetectable period of the sound source information. . That is, it is determined whether or not the following equations (7) and (8) hold. T ₀ ≦ T _i ≦ T ₁ , and T ₂ ≦ T _j ≦ T ₃ (7) θ _t −α <θ ′ <θ _t + α (8) where α is an error range of the detected prediction azimuth. When the above equations (7) and (8) are satisfied, the data combination determination device 6 combines the integration result data of the target A and the integration result data of the target B as in the above equation (4).

[0025]

(Equation 3)

As described above, according to the second embodiment, the following effects can be obtained by using the spectrum correlation determination device 4A. (1) Resource Saving by Reducing the Amount of Information In the case of the first embodiment, when obtaining the certainty factor of the sound source from the frequency information, the data of the information of the sound source must be held in advance. On the other hand, in the second embodiment, since the sound source is not specified, there is no need to hold data, and resources can be saved. (2) Reduction of Calculation Amount In the case of the first embodiment, the process of calculating the certainty factor of the sound source based on the Dempster-Scheffer theory depends on the amount of information information of the sound source described above in addition to the detected frequency information. Although the amount of calculation increases at an accelerated rate, the calculation for obtaining the spectrum correlation depends only on the combination of the spectra, so that the amount of calculation can be reduced as compared with the case of the first embodiment.

In the apparatus configuration shown in FIG. 4, an example is shown in which the data combination determination device 6 is applied to the signal processing device 1, the signal detection and tracking device 2, the data integration device 3, the spectrum correlation device 4A and the target motion analysis device 5. Since the information from the target motion analysis device 5 is the predicted orientation of the target that has once been lost, the data from the data integration device 3 is used instead of the target motion analysis device 5, and similarly, the target T seconds The same effect can be obtained by using a device that predicts the later appearance direction.

[0028]

As described above, according to the present invention, during the observation period, the frequency of the acoustic signal from the sound source received through the acoustic sensor having directivity is analyzed, and Signal strength data is obtained, sound source information is detected and tracked based on the data, direction information and frequency information determined as the same sound source are integrated, and the entity name of the sound source and its certainty are obtained based on the integrated frequency information. At the same time, a target motion analysis based on azimuth and frequency information obtained in a time series is used to obtain a detection prediction azimuth range and a prediction accuracy of a sound source after a predetermined time, and the sound source information detected so far during the observation period is once obtained. If there is sound source information that has become undetectable and can be detected again within a certain period thereafter, the two sound source information detected during the observation period before and after the non-detection period of the sound source information, respectively. (1) Two entity names obtained from the sound source information match during an observation period before and after the sound source information undetectable period, and a certainty factor for each entity name Must be equal to or greater than the specified value. (2) The azimuth component of the sound source information detected in the observation period after the undetectable period is included in the detection azimuth range after a predetermined time of the sound source obtained in the observation period before the undetectable period of the sound source information. . When both the above conditions (1) and (2) are satisfied, the two sound source information detected in the observation periods before and after the non-detection period of the sound source information are temporally combined. Compared with the conventional method, erroneous coupling is suppressed, the coupling accuracy is improved, the workload of the operator is reduced by automation, and the data processing performance of target motion analysis is improved.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a data time-series combination device according to Embodiment 1 of the present invention.

FIG. 2 is an operation explanatory diagram in which input / output signals are entered in the device configuration diagram of FIG. 1;

FIG. 3 is a diagram showing azimuth information of two targets arranged in time series.

FIG. 4 is a configuration diagram of a data time-series combination device according to Embodiment 2 of the present invention.

FIG. 5 is an explanatory diagram of a conventional detection target combining operation.

FIG. 6 is an explanatory diagram of a problem in the conventional target combination.

[Description of Signs] 1 signal processing device 2 signal detection and tracking device 3 data integration device 4 target class identification device 4A spectrum correlation determination device 5 target motion analysis device 6 data connection determination device

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Hiroshi Manzoji F-term (reference) in Oki Electric Industry Co., Ltd. 1-7-12 Toranomon, Minato-ku, Tokyo 5J083 AA05 AB10 AC17 AC40 AD18 AE07 AF15 BC01 BE41 BE45 BE60

Claims (6)

[Claims] During an observation period, an acoustic signal from a sound source received through an acoustic sensor having directivity is subjected to frequency analysis to obtain signal strength data for each receiving direction and frequency, and based on the data, Detects and tracks sound source information, integrates direction information and frequency information determined to be the same sound source, obtains the entity name of the sound source and its certainty based on the integrated frequency information, and obtains the direction and frequency obtained in time series. The target motion analysis based on the information to determine the detection prediction azimuth range and the prediction accuracy after a predetermined time of the sound source, the sound source information that has been detected so far during the observation period becomes temporarily undetectable,
If there is sound source information that can be detected again within a certain period thereafter, whether or not the two sound source information detected respectively in the observation periods before and after the undetectable period of the sound source information can be temporally combined is determined. (1) The two entity names obtained from the sound source information match during the observation period before and after the sound source information cannot be detected period, and the certainty factor for each entity name is both a specified value or more. (2) The azimuth component of the sound source information detected in the observation period after the undetectable period is included in the detection azimuth range after a predetermined time of the sound source obtained in the observation period before the undetectable period of the sound source information. When both of the above-mentioned conditions (1) and (2) are satisfied, the two pieces of sound source information respectively detected in the observation periods before and after the non-detection period of the sound source information are temporally combined. Data time series Coupling methods. 2. The target motion analysis based on the azimuth and frequency information obtained in the time series includes performing a target motion analysis process using a nonlinear least squares method or a Kalman filter. Data time series join method. 3. The target motion analysis based on the azimuth and frequency information obtained in the time series, the frequency information that has been detected for the longest time or the frequency with the highest signal strength is selected as the representative frequency, and the nonlinear minimum is selected. 2. The data time series combining method according to claim 1, wherein a target motion analysis process using a square method or a Kalman filter is performed. 4. During the observation period, an acoustic signal from a sound source received through an acoustic sensor having directivity is subjected to frequency analysis to obtain signal strength data for each receiving direction and frequency, and based on the data, Detects and tracks sound source information, integrates azimuth information and frequency information determined to be the same sound source, determines whether or not there is a frequency spectrum correlation with the integrated frequency information, and determines a sound source and a time that are recognized as having a correlation. While obtaining the combination information with, the target motion analysis based on the azimuth and frequency information obtained in time series to obtain the detection prediction azimuth range and the prediction accuracy after a predetermined time of the sound source, and have been detected so far during the observation period. If the sound source information was once undetectable, and then there is sound source information that can be detected again within a certain period after that, the sound source information is not detected during the observation period before and after the undetectable period of the sound source information. Conditions for determining whether the two detected sound source information can be temporally combined include: (1) two integrated frequency information obtained from the sound source information in observation periods before and after an undetectable period of the sound source information, respectively. (2) an observation period after a non-detection period within a detection prediction azimuth range after a predetermined time period of the sound source obtained in an observation period before the non-detection period of the sound source information. When both of the above-described determination conditions (1) and (2) that the azimuth component of the detected sound source information is included, the two detections are performed during the observation periods before and after the non-detection period of the sound source information. A data time-series combining method characterized by temporally combining sound source information. 5. The target motion analysis based on the azimuth and frequency information obtained in a time series, the target motion analysis processing using a nonlinear least squares method or a Kalman filter is performed. Data time series join method. 6. The target motion analysis based on the azimuth and frequency information obtained in the time series, selects the frequency detected for the longest time or the frequency with the highest signal strength as the representative frequency for the frequency information, 5. The data time series combining method according to claim 4, wherein a target motion analysis process using a square method or a Kalman filter is performed.