Tracking device, tracking system, tracking method, and tracking method program

Abstract

To obtain a tracking device or the like capable of continuously tracking an object regardless of a state of an SN ratio.SOLUTION: A tracking device includes: a likelihood generation section 111 for executing processing to generate likelihood functions from observation value data of a signal related to a tracking object; a Bayes estimation calculation section 112 for generating a temporary posterior probability distribution from the likelihood functions; a first test quantity calculation section 113 for generating a first test quantity from the temporary posterior probability distribution; a Markov update section 114 to serve as a prior probability distribution generation section for generating a prior probability distribution related to an orientation of the tracking object; a second test quantity calculation section 115 for generating a second test quantity from the prior probability distribution; a likelihood input determination section 116 for determining one of the prior probability distribution and the temporary posterior probability distribution in the processing before one cycle as a final posterior probability distribution; a test quantity control section 117 for defining one of the first test quantity and the second test quantity as the final test quantity based on the determined final posterior probability distribution; and a detection determination section 118 for determining presence / absence of the tracking object based on the final test quantity.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] This technology relates to a tracking device, a tracking system, a tracking method, and a tracking method program that track a target based on a signal, and in particular, to ensuring continuity of tracking. [Background technology]

[0002] Conventionally, there is known a tracking device that estimates a state quantity indicating the presence of a tracked target in a state quantity space based on a signal emitted by the object to be tracked in a noisy environment, and tracks the target based on the estimation. For example, the tracking device performs tracking by using a sound wave emitted by the tracked target as a signal and estimating the direction of arrival of the sound wave and the time change in frequency, etc. The state quantity indicating the presence of a tracked target corresponds to the direction of arrival of the sound wave and the frequency of the sound wave (for example, see Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] REBethel and GJParas, “A PDF Multitarget Tracker”, IEEE Transactions on Aerospace and Electronic Systems, vol. 30, NO. 2, April 1994, p. 386-403 Summary of the Invention [Problem to be solved by the invention]

[0004] In this case, if the signal-to-noise ratio (hereinafter referred to as the SNR) is high, the frequency distribution of the likelihood of signal plus noise will be separated from the distribution of the likelihood of noise only. As a result, by performing Bayesian estimation, the posterior probability of the signal direction can be improved over time.

[0005] However, when the signal-to-noise ratio is low, there is an intersection between the signal-to-noise likelihood distribution and the noise-only distribution, so there are times when the likelihood value for the signal direction is high and times when the likelihood value for the signal direction is low, even if the frequency distribution of the likelihood is similar.

[0006] As a result, when tracking a target based on the posterior probability, the posterior probability drops significantly at times when the likelihood becomes low, making it impossible to identify the signal related to the tracking target and its direction of arrival. As such, the occurrence of variations in likelihood can lead to the tracking target being lost.

[0007] Therefore, there has been a demand for a tracking device that can continuously track a target regardless of the signal-to-noise ratio. [Means for solving the problem]

[0008] The tracking device according to the present disclosure comprises: A tracking device that determines the presence or absence of a tracking target for each period of processing observed value data obtained by observing a signal related to the tracking target, From observed data In multiple directions Calculate the likelihood and the direction are the parameter variables a likelihood generation unit that performs processing to generate a likelihood function; a Bayesian estimation calculation unit that performs processing to generate a tentative posterior probability distribution from the likelihood function; a first calibration metric calculation unit that performs processing to generate a first calibration metric from the tentative posterior probability distribution; and a final posterior probability distribution of a likelihood input determination unit that performs a process of determining likelihood; Based on the final posterior probability distribution determined by the likelihood input determination unit, a prior probability distribution generation unit that performs processing to generate a prior probability distribution related to the direction of a tracking target; generated by the prior probability distribution generator a second calibration metric calculation unit that performs processing to generate a second calibration metric from the prior probability distribution; The results of the comparison based on the first and second quantifiers showed that a calibration amount control unit that performs processing to determine either the first calibration amount or the second calibration amount as a final calibration amount; and A certain period A detection determination unit performs processing to determine whether or not a tracking target is present, the likelihood input determination unit determines, based on the first validation quantity and the final validation quantity determined by the validation quantity control unit in the processing one cycle before the given cycle, either the provisional posterior probability distribution or the prior probability distribution generated by the prior probability distribution generation unit in the processing one cycle before the given cycle, as the final posterior probability distribution; It is something.

[0009] In addition, the tracking system according to this disclosure comprises a receiver array having a plurality of receivers and a signal conversion device that converts signals received by the receivers into digital data to generate observation value data, and the tracking device described above.

[0010] The tracking method according to the present disclosure also includes: The process is performed periodically. A tracking method performed by a data processing device that tracks a tracking target, comprising: How to do it teeth, In a certain period From observed data In multiple directions Calculate the likelihood and the direction are the parameter variables generating a likelihood function; generating a tentative posterior probability distribution from the likelihood function; and generating a first estimator from the tentative posterior probability distribution; A certain period Final posterior probability distribution of decision and based on the final posterior probability distribution, generating a prior probability distribution relating to the direction of the tracking target; and generating a second calibration metric from the prior probability distribution; The results of the comparison based on the first and second quantifiers showed that determining either the first calibration amount or the second calibration amount as a final calibration amount; and based on the final calibration amount, A certain period and a step of determining whether or not there is a tracking target. the step of determining the final posterior probability distribution of a certain cycle includes determining, based on the first calibration quantifier and the final calibration quantifier in the process one cycle before the certain cycle, either the tentative posterior probability distribution or the prior probability distribution in the process one cycle before the certain cycle as the final posterior probability distribution of the certain cycle; It is something.

[0011] Further, a tracking method program according to the present disclosure is a tracking method program for tracking a tracking target based on observation value data obtained from a signal related to the tracking target, In a certain period From observed data In multiple directions Calculate the likelihood and the direction are the parameter variables a step of generating a likelihood function, a step of generating a tentative posterior probability distribution from the likelihood function, and a step of generating a first statistic from the tentative posterior probability distribution; A certain period Final posterior probability distribution of decision and performing a process of: a step of generating a prior probability distribution related to the direction of the tracking target; and a step of generating a second calibration metric from the prior probability distribution. The results of the comparison based on the first and second quantifiers showed that performing a process in which either the first calibration amount or the second calibration amount is set as a final calibration amount; and determining, based on the final calibration amount, A certain period and causing a computer to perform a process of determining whether or not a tracking target is present. wherein the step of performing processing to determine the final posterior probability distribution of a certain period determines, based on the first calibration value and the final calibration value in processing one period before the certain period, either the tentative posterior probability distribution or the prior probability distribution in processing one period before the certain period as the final posterior probability distribution of the certain period. It is something. [Effects of the Invention]

[0012] According to this disclosure, a first validation quantifier calculation unit generates a first validation quantifier from the tentative posterior probability distribution generated by the Bayesian estimation calculation unit. Furthermore, a second validation quantifier calculation unit generates a second validation quantifier from the prior probability distribution generated by the Markov update unit. The likelihood input determination unit then sets either the tentative posterior probability distribution or the prior probability distribution as the final posterior probability distribution, and the validation quantifier control unit sets either the first validation quantifier or the second validation quantifier as the final validation quantifier. By using the prior probability distribution and the second validation quantifier obtained by processing one cycle prior instead of the tentative posterior probability distribution and the first validation quantifier generated using a likelihood function with a low likelihood, a sudden decrease in the validation quantifier can be prevented, and the loss of the tracking target can be prevented. Furthermore, the tracking device can improve tracking continuity by preventing observation value data at times with low likelihood from affecting the next cycle or later. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a diagram showing the configuration of a tracking system centered around a tracking device 100 according to a first embodiment. [Figure 2] FIG. 2 is a diagram illustrating a processing flow in a data processing device 110 according to the first embodiment. [Figure 3] FIG. 10 is a diagram showing the relationship between the likelihood function lt(θ), the prior probability distribution Pprior t|t−1, and the posterior probability distribution Pposterior t(θ) in a situation where the SNR between the signal related to the target and the noise is high. [Figure 4] FIG. 10 is a diagram showing the relationship between the likelihood function lt(θ), the prior probability distribution Pprior t|t−1, and the posterior probability distribution Pposterior t(θ) when the SNR between the target signal and noise is low and the likelihood of the signal direction is large. [Figure 5] FIG. 10 is a diagram showing the relationship between the likelihood function lt(θ), the prior probability distribution Pprior t|t−1, and the posterior probability distribution Pposterior t(θ) when the SNR between the target signal and noise is low and the likelihood of the signal direction is small. [Figure 6] FIG. 3 is a diagram illustrating the effect of the tracking system according to the first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] Hereinafter, a tracking device according to an embodiment will be described with reference to the drawings. In the following drawings, components with the same reference numerals are identical or equivalent, and are common throughout the following embodiments. Furthermore, the size relationships between components in the drawings may differ from those in reality. Furthermore, the configurations of components shown throughout the specification are merely examples and are not limited to the configurations described in the specification. Not all of the devices described in the specification may be included. In particular, the combinations of components are not limited to the combinations in each embodiment, and components described in other embodiments may be applied to other embodiments. Furthermore, when multiple similar devices are distinguished by subscripts, the reference numerals, subscripts, etc. may be omitted if there is no need to distinguish or identify them.

[0015] Embodiment 1 FIG. 1 is a diagram showing the configuration of a tracking system centered on a tracking device 100 according to a first embodiment. In explaining the tracking system according to the first embodiment, an acoustic signal generated by a sound wave will be taken as an example of a signal used to track a target to be tracked. Hereinafter, the target to be tracked will be referred to as a target. The acoustic signal includes sound waves that are signals obtained from the target that is the sound source, as well as sound waves that are noise. Here, the explanation will be given assuming that there is one target (sound source) that the tracking system according to the first embodiment tracks.

[0016] The tracking system according to the first embodiment includes a tracking device 100, a microphone array 200, and a communication line 300. As shown in FIG. 1, the microphone array 200 is a receiver array in which a plurality of microphones 210-1 to 210-n, which convert received acoustic signals into electrical signals, are arranged in an arbitrary shape. The microphone array 200 also includes a signal conversion device 220 and a signal transmission device 230. The signal conversion device 220 periodically performs sampling and encoding processes on the acoustic signals received by each microphone 210 to convert them into observation value data, which is digitalized data to be processed, and performs data generation processing. The signal transmission device 230 transmits a signal including the observation value data.

[0017] The communication line 300 is a line that connects the microphone array 200 and the tracking device 100 for communication. Here, the communication line 300 is assumed to be a dedicated line for communication, but is not limited to this. For example, a signal may be sent from the microphone array 200 to the tracking device 100 via a commonly used electric communication line. The communication line 300 may be wired or wireless.

[0018] Tracking device 100 is a device that performs processing related to target tracking based on observation data. Tracking device 100 in embodiment 1 includes data processing device 110, storage device 120, signal input device 130, and signal output device 140. Based on a signal sent from microphone array 200 via communication line 300, signal input device 130 converts data in the signal into observation data that can be processed by data processing device 110. Furthermore, signal output device 140 generates a signal including data on the estimated direction and detection determination result processed by data processing device 110, and transmits the signal to an external device (not shown). Here, in embodiment 1, tracking device 100 includes signal input device 130 and signal output device 140, but these may be devices other than tracking device 100.

[0019] The data processing device 110 included in the tracking device 100 of the first embodiment generates data on an estimated direction and a detection determination result based on observation data related to an acoustic signal. The data processing device 110 includes a likelihood generation unit 111, a Bayesian estimation calculation unit 112, a first calibration quantifier calculation unit 113, a Markov update unit 114, a second calibration quantifier calculation unit 115, a likelihood input determination unit 116, a calibration quantifier control unit 117, a detection determination unit 118, and a direction estimation unit 119. Each unit in the data processing device 110 performs processing such as calculation and determination, and generates various data for processing by the data processing device 110.

[0020] In the data processing device 110, the likelihood generating unit 111 generates likelihoods in a plurality of directions based on the observation value data relating to the acoustic signals received by the microphone array 200, and calculates a likelihood function l with the direction θ as a parameter variable. t Here, t represents time (t=1, 2, ...), and the period from time t to time t+1 is the processing cycle of the data processing device 110. The likelihood generation unit 111 performs integration to calculate the likelihood function l t (θ) is calculated. For example, the processing period is set to a time period during which the likelihood generation unit 111 can perform integration. The tracking device 100 of the first embodiment processes observation value data at a certain time t. The likelihood generation unit 111 forms directivities in a plurality of directions simultaneously by, for example, beamforming. The likelihood generation unit 111 then arranges and processes the power values ​​of the sound in each direction, and calculates the obtained data using the likelihood function l t Here, the likelihood generator 111 calculates a likelihood function l using beamforming. t (θ), but is not limited to this. For example, by assuming a distribution of the sound source direction, it is possible to generate a likelihood function l related to the vicinity of the instantaneous estimated direction of the sound source. t (θ) to calculate other likelihood functions l t Techniques are available to generate (θ).

[0021] The Bayesian estimation calculation unit 112 calculates the likelihood function l related to the processing of the likelihood generation unit 111. t(θ) and the prior probability distribution P prior t|t-1 (θ) is used to calculate the provisional posterior probability distribution P posterior t (θ) is calculated by the Bayesian estimation calculation unit 112. posterior t (θ) is used as data in the processing performed by the direction estimation unit 119, the first calibration quantity calculation unit 113, and the likelihood input determination unit .

[0022] Here, the prior probability distribution P prior t|t-1 (θ) is the prior probability distribution P at one period before time t (time t−1) stored in the prior probability buffer unit 121 of the storage device 120. prior t|t-1 (θ). Here, the prior probability distribution P prior 1|0 Generally, if the initial value of (θ) is not set in advance, it is set to a value that has equal probability for all target directions. On the other hand, if it is set in advance, the prior probability distribution P prior t|t-1 The initial value of (θ) is the pre-defined prior probability distribution P prior 1|0 The value is set to an arbitrary distribution based on (θ).

[0023] The first calibration quantifier calculation unit 113 calculates the tentative posterior probability distribution P posterior t Then, the first calibration quantity calculation unit 113 calculates the maximum value in the posterior probability from (θ). t Here, the maximum value in the posterior probability is the first test metric TS1 t However, the present invention is not limited to this example. For example, if it is possible to determine whether or not a target that is a sound source exists, such as by adding a certain range centered on the maximum value in the posterior probability, the first calibration quantity TS1 tThe first calibration quantity TS1 determined by the first calibration quantity calculation unit 113 can be expressed as follows: t is used as data in the processing of the likelihood input determination unit 116 and the validation quantity control unit 117.

[0024] The likelihood input determination unit 116 calculates the tentative posterior probability distribution P posterior t (θ), the first calibration quantity TS1 determined by the first calibration quantity calculation unit 113 t , the prior probability distribution P prior t|t-1 (θ), the final test amount TS at one cycle before (time t-1) stored in the final test amount buffer unit 122, and the likelihood input decision threshold Th_ΔTS stored in the likelihood input decision threshold storage unit 123. Then, the final posterior probability distribution P posterior.Det t (θ) is determined, and the value of the likelihood input determination flag is set.

[0025] The Markov update unit 114 updates the final posterior probability distribution P posterior.Det t Based on (θ), the prior probability distribution P prior t+1|t The Markov update unit 114 calculates and generates the final posterior probability distribution P posterior.Det t The prior probability distribution P prior t+1|t (θ) is calculated by the Markov update unit 114. prior t+1|t (θ) is used in the processing of the second calibration quantity calculation unit 115. In addition, the prior probability distribution P prior t+1|t (θ) is stored in the prior probability buffer unit 121, which will be described later. The prior probability distribution P prior t+1|t(θ) is the prior probability distribution P at one period before (time t) in the processing at one period after time t (time t+1) in the Bayesian estimation calculation unit 112 and the likelihood input determination unit 116. prior t|t-1 (θ) is used.

[0026] The second calibration quantifier calculation unit 115 calculates the prior probability distribution P at time t calculated by the Markov update unit 114 through calculation processing. prior t+1|t From (θ), the second calibration metric TS2 t The second calibration quantity TS2 determined by the second calibration quantity calculation unit 115 is t is used as data in the processing of the calibration quantity control unit 117. Here, the second calibration quantity calculation unit 115 calculates the prior probability distribution P prior t+1|t From (θ), the maximum value in the prior probability is the second test TS2 t However, it is not limited to this. For example, if it is possible to determine whether or not a target that is a sound source exists, such as by adding up a certain range centered on the maximum value in the prior probability, the second calibration quantity TS2 t It can be said that:

[0027] The calibration quantity control unit 117 calculates the first calibration quantity TS1 generated by the first calibration quantity calculation unit 113 based on the likelihood input determination flag from the likelihood input determination unit 116. t and the second calibration quantity TS2 generated by the second calibration quantity calculation unit 115 t The final validation amount TS determined by the validation amount control unit 117 is used as data in the processing by the detection determination unit 118. The final validation amount TS is also stored as data in the final validation amount buffer unit 122. The final validation amount TS stored in the final validation amount buffer unit 122 is used as data in the processing performed by the likelihood input determination unit 116 one cycle later (at time t+1).

[0028] The detection determination unit 118 performs a determination process to determine whether or not there is a target that serves as a sound source, based on the final verification amount TS determined by the verification amount control unit 117.

[0029] The direction estimation unit 119 also calculates the tentative posterior probability distribution P posterior t Based on (θ), direction estimation processing is performed to estimate the direction of the sound source (target) from which the acoustic signal is emitted. Here, the direction estimation processing performed by the direction estimation unit 119 is processing performed using techniques such as MAP estimation and weighted averaging.

[0030] Furthermore, the storage device 120 included in the tracking device 100 according to the first embodiment temporarily or long-term stores data used for processing by the data processing device 110. The storage device 120 according to the first embodiment particularly includes a prior probability buffer unit 121, a final validation amount buffer unit 122, and a likelihood input determination threshold storage unit 123.

[0031] The prior probability buffer unit 121 stores the prior probability distribution P prior t+1|t (θ) for one period. As described above, the prior probability distribution P prior t+1|t (θ) is the prior probability distribution P prior t|t-1 (θ) is used.

[0032] The final calibration amount buffer unit 122 stores the final calibration amount TS at time t determined by the calibration amount control unit 117 for one period.

[0033] The likelihood input decision threshold storage unit 123 stores the likelihood input decision threshold Th_ΔTS as data, which is used when the verification quantity control unit 117 performs processing. The likelihood input decision threshold Th_ΔTS is a parameter that is set in advance. The likelihood input decision threshold Th_ΔTS is set taking into consideration the trade-off between the continuity of signals related to targets and the continuity of false alarms. The smaller the likelihood input decision threshold Th_ΔTS, the better the continuity of signals related to targets, but when a false alarm occurs, the false alarm also tends to continue and reliability tends to decrease. On the other hand, the larger the likelihood input decision threshold Th_ΔTS, the less likely a false alarm will continue, but the less likely the signal will continue. For this reason, for example, with regard to the continuity of false alarms, a standard such as "the probability that a false alarm will continue for 10 seconds or more is 1% or less" is set, and the smallest value that meets the standard is set as the likelihood input decision threshold Th_ΔTS.

[0034] Here, the data processing device 110 of the tracking device 100 is typically configured with a device that performs control and arithmetic processing, such as a computer centered around a CPU (Central Processing Unit).The data processing device 110 typically executes a pre-programmed signal processing procedure performed by each unit to realize the processing of each unit.Here, for example, the storage device 120 stores the program data.However, this is not limited to this, and each unit may be configured with a separate dedicated device (hardware).

[0035] The storage device 120 also includes a volatile storage device (not shown) such as random access memory (RAM) that can temporarily store data, and a non-volatile auxiliary storage device (not shown) such as a hard disk drive. The non-volatile auxiliary storage device may be a solid state drive, a flash memory that can store data for a long period of time, or the like.

[0036] FIG. 2 is a diagram illustrating a processing flow in the data processing device 110 according to the first embodiment. Next, the processing operation of the data processing device 110 according to the first embodiment will be described. As described above, the processing of the data processing device 110 at time t will be described here. The likelihood generating unit 111 generates likelihoods in a plurality of directions based on the observation value data relating to the acoustic signal received by the microphone array 200, as described above, and calculates a likelihood function l t (θ) is calculated and generated (step S1).

[0037] The Bayesian estimation calculation unit 112 calculates the likelihood function l related to the processing of the likelihood generation unit 111. t (θ) and the prior probability distribution P prior t|t-1 Based on (θ), the tentative posterior probability distribution P posterior t (θ) is calculated and generated (step S2). At this time, the Bayesian estimation calculation unit 112 calculates and generates a tentative posterior probability distribution P posterior t (θ) is calculated. Here, θ' is a variable related to the direction when calculating the summation (summation). Therefore, the denominator in equation (1) is the sum of the products of the likelihood and the prior probability one cycle before in all directions.

[0038]

number

[0039] The first calibration quantifier calculation unit 113 calculates the tentative posterior probability distribution P posterior t From (θ), the first calibration metric TS1 t is determined and generated (step S3).

[0040] The likelihood input determination unit 116 calculates the final test quantity TS at the previous cycle (time t-1) stored in the final test quantity buffer unit 122 and the first test quantity TS1 determined by the first test quantity calculation unit 113 based on the following equation (2): tCalculate the difference ΔTS, which is the difference from that (step S4).

[0041] [Number]

[0042] Then, as shown in the following equation (3), the likelihood input determination unit 116 compares the likelihood input determination threshold Th_ΔTS stored in the likelihood input determination threshold storage unit 123 with the difference ΔTS, and determines the final posterior probability distribution P posterior.Det t (θ) at time t, and sets the value of the likelihood input determination flag flag. When the likelihood input determination unit 116 determines that ΔTS < Th_ΔTS, the temporary posterior probability distribution P posterior t (θ) generated by the Bayesian estimation calculation unit 112 through arithmetic processing is determined as the final posterior probability distribution P posterior.Det t (θ), and the value of the likelihood input determination flag flag is set to 1. On the other hand, when the likelihood input determination unit 116 determines that ΔTS ≥ Th_ΔTS, the prior probability distribution P prior t|t-1 (θ) one cycle before stored in the prior probability buffer unit 121 is determined as the final posterior probability distribution P posterior.Det t (θ), and the value of the likelihood input determination flag flag is set to 0 (step S5).

[0043] [Number]

[0044] Here, the processing of the Markov update unit 114 will be described. The Markov update unit 114 calculates and generates the prior probability distribution P posterior.Det t (θ) at time t based on the final posterior probability distribution P prior t+1|t (θ) determined by the likelihood input determination unit 116 (step S6). The Markov update unit 114 stores the prior probability distribution P prior t+1|tThe second calibration quantifier calculation unit 115 stores the prior probability distribution P (θ) at time t calculated by the Markov update unit 114. prior t+1|t Based on (θ), the second calibration metric TS2 t is determined and generated (step S7).

[0045] As shown in the following equation (4), if the likelihood input determination flag set by the likelihood input determination unit 116 is set to a value of 1, the calibration quantity control unit 117 calculates the first calibration quantity TS1 t On the other hand, if the value of the likelihood input determination flag is 0, the second calibration quantity TS2 t is determined as the final calibration amount TS (step S8).

[0046]

number

[0047] The detection determination unit 118 compares the final calibration amount TS determined by the calibration amount control unit 117 with a predetermined final detection amount threshold Th_TS to determine whether a target that is a sound source has been detected, and generates a detection determination result (step S9). A signal including data on the detection determination result is output by the signal output device 140 as a determination result signal.

[0048] On the other hand, the direction estimation unit 119 calculates the tentative posterior probability distribution P posterior t A direction estimation process is performed based on (θ) to generate an estimated direction (step S10). A signal including data relating to the estimated direction of the target is output by the signal output device 140 as an estimated direction signal.

[0049] Figure 3 shows the likelihood function l in the situation where the signal-to-noise ratio of the target is high. t (θ), the prior probability distribution P prior t|t-1 and the posterior probability distribution P posterior t3 is a diagram showing the relationship between the likelihood function l and the likelihood function θ. As shown in FIG. 3, in a situation where the SNR is high, the frequency distribution of the likelihood of signal plus noise is separated from the distribution of the likelihood of noise only. t (θ) and the prior probability distribution P prior t|t-1 The posterior probability distribution P calculated based on posterior t The value of the calibration quantity obtained from (θ) is large, and the presence of the target and the signal direction are easily obtained.

[0050] Figure 4 shows the likelihood function l when the signal-to-noise ratio of the target signal is low and the likelihood of the signal direction is large. t (θ), the prior probability distribution P prior t|t-1 and the posterior probability distribution P posterior t 5 shows the relationship between the likelihood function l and the target direction when the signal-to-noise ratio is low and the likelihood of the signal direction is small. t (θ), the prior probability distribution P prior t|t-1 and the posterior probability distribution P posterior t 4 and 5 show the relationship between the posterior probability distribution P posterior t In (θ), it becomes difficult to estimate the signal direction. Also, the value of the calibration index becomes small, which may cause the target to be lost.

[0051] 6 is a diagram illustrating the effect of the tracking system according to the first embodiment. posterior t First calibration metric TS1 by (θ) t If the detection determination unit 118 determines the presence or absence of a target only by the first calibration quantity TS1, when the likelihood of the target being present in the direction is low as shown in FIG. 5, as shown in FIG. 6(a),t becomes lower than the final detection amount threshold Th_TS, and the target is lost. On the other hand, the prior probability distribution P prior t|t-1 The second calibration metric TS2 obtained from (θ) t Then, the likelihood function l with low likelihood t Therefore, as shown in Figure 6(b), the final calibration amount TS (TS in Figure 6(b)) of the previous cycle is not based on (θ). t-1 ) becomes smaller, and the first calibration metric TS1 t Even if the distance is small, it can prevent the target from being lost.

[0052] As described above, according to the tracking device 100 of the first embodiment, the Bayesian estimation calculation unit 112 calculates the tentative posterior probability distribution P posterior t (θ), and the first calibration quantity calculation unit 113 calculates the first calibration quantity TS1 t Furthermore, the Markov update unit 114 generates the prior probability distribution P prior t+1|t (θ), and the second calibration quantity calculation unit 115 calculates the second calibration quantity TS2 t Then, the likelihood input determination unit 116 generates the final test amount TS at time t−1 and the first test amount TS1 at time t. t The difference ΔTS between the two is compared with the likelihood input decision threshold Th_ΔTS, and the provisional posterior probability distribution P posterior t (θ) or the prior probability distribution P at time t-1 prior t|t-1 (θ) is used as the final posterior probability distribution P posterior.Det t (θ). Based on the determination of the likelihood input determination unit 116, the calibration quantity control unit 117 calculates the first calibration quantity TS1 t or the second calibration metric TS2 t In this way, the posterior probability distribution P calculated by the Bayesian estimation calculation unit 112 is used as the final test value TS. posterior t First calibration quantity TS1 related to (θ) t If is less than the reference likelihood input decision threshold Th_ΔTS, the prior probability distribution P prior t|t-1(θ) is the final posterior probability distribution P posterior.Det t (θ), and the second calibration metric TS2 t This prevents a sudden drop in the test quantity. posterior.Det t (θ) is the prior probability distribution P prior t|t-1 (θ) is adopted, and the second calibration value TS2 is used as the final calibration value TS. t By adopting this, the likelihood function l at time t t (θ) will not be reflected in the data generated in the subsequent cycles. Therefore, the likelihood function l t The data that has been subjected to arithmetic processing using (θ) does not affect the next period and thereafter. Therefore, the tracking device 100 of the tracking system according to the first embodiment can prevent the target from being lost and improve the continuity of tracking even if the S / N ratio is low.

[0053] Embodiment 2 As an embodiment other than the above-described first embodiment, there is the following form. For example, in the tracking device 100 of the above-described first embodiment, the likelihood input decision unit 116 of the data processing device 110 calculates the final validation amount TS in the previous period and the first validation amount TS1. t Based on the difference ΔTS between posterior.Det t (θ) is determined. However, the present invention is not limited to this. For example, the likelihood input determination unit 116 determines the final verification amount TS in the previous cycle and the first verification amount TS1 t The ratio TSA is calculated by comparing the ratio TSA with a preset likelihood input decision threshold Th_TSA to obtain the final posterior probability distribution P posterior.Det t (θ). In this way, the first calibration metric TS1 generated from the observed value data may be determined as t It is sufficient if it can determine whether there is a sudden drop in the

[0054] In the first embodiment described above, the direction estimation unit 119 calculates the tentative posterior probability distribution P posteriort The direction estimation process was performed based on (θ) to generate the estimated direction. posterior t By generating an estimated direction based on (θ), the tracking device 100 can perform tracking that reflects the likelihood of the acoustic signal at time t in the direction estimation. However, this is not limited to this. The direction estimation unit 119 calculates the final posterior probability distribution P posterior.Det t An azimuth estimation process may be performed based on (θ) to generate an estimated azimuth.

[0055] In the first embodiment, the Markov update unit 114 generates the prior probability distribution P prior t+1|t (θ) was calculated and generated. However, this is not limited to this. The prior probability distribution P prior t+1|t (θ) is a probability distribution assumed before (before) processing is performed by the Bayesian estimation calculation unit 112. For this reason, if the direction assumed as prior knowledge can be expressed as a probability distribution, the prior probability distribution generation unit can regard it as a prior probability distribution and generate the prior probability distribution.

[0056] In the first embodiment described above, the tracking device 100 tracks a target based on acoustic signals received by the microphones 210 of the microphone array 200. Here, the microphone array 200 of the first embodiment is assumed to be mounted on a passive sonar, but is not limited to this. It can also be applied to an active sonar that emits acoustic waves and uses the reflected acoustic waves as an acoustic signal. It can also be applied to signals obtained by RADER (Radio Detecting and Ranging), LIDAR (Light Detection and Ranging), etc.

[0057] Furthermore, the tracking device 100 in the tracking system statistically ensures the S / N ratio, but can track a target not only when the S / N ratio drops momentarily due to noise variations, but also when the S / N ratio continues to drop for a certain period of time, for example, due to interference. [Explanation of symbols]

[0058] 100 Tracking device 110 Data processing device 111 Likelihood generation unit 112 Bayesian Estimation Calculation Unit 113 First calibration quantity calculation unit 114 Markov Update Unit 115 Second calibration quantity calculation unit 116 Likelihood input decision unit 117 Calibration quantity control section 118 Detection and judgment unit 119 Orientation estimation part 120 Storage device 121 Prior probability buffer section 122 Final calibration amount buffer section 123 Likelihood input decision threshold memory unit 130 Signal input device 140 Signal output device 200 microphone array 210, 210-1 to 210-n microphones 220 Signal conversion device 230 Signal Transmitting Device 300 communication lines

Claims

1. A tracking device that determines the presence or absence of a tracking target at each period in which observation value data obtained by observing a signal related to the tracking target is processed, comprising: a likelihood generation unit that calculates likelihoods in a plurality of directions from the observation value data in a certain period and generates a likelihood function with the directions as parameter variables; a Bayesian estimation calculation unit that performs processing to generate a tentative posterior probability distribution from the likelihood function; a first calibration metric calculation unit that performs processing to generate a first calibration metric from the provisional posterior probability distribution; a likelihood input determination unit that performs processing to determine a final posterior probability distribution; a prior probability distribution generation unit that generates a prior probability distribution related to the direction of the tracking target based on the final posterior probability distribution determined by the likelihood input determination unit; a second calibration metric calculation unit that performs processing to generate a second calibration metric from the prior probability distribution generated by the prior probability distribution generation unit; a calibration amount control unit that performs processing to determine either the first calibration amount or the second calibration amount as a final calibration amount based on a comparison result based on the first calibration amount and the second calibration amount; a detection determination unit that performs processing to determine whether or not the tracking target exists in the certain period based on the final verification amount; Equipped with the likelihood input determination unit determines, based on the first validation quantity and the final validation quantity determined by the validation quantity control unit in the processing one cycle before the certain cycle, either the provisional posterior probability distribution or the prior probability distribution generated by the prior probability distribution generation unit in the processing one cycle before the certain cycle, as the final posterior probability distribution. Tracking device.

2. a final calibration amount buffer unit that stores the final calibration amount determined by the calibration amount control unit as data; The likelihood input determination unit 2. The tracking device according to claim 1, wherein a process of determining the final posterior probability distribution is performed based on a difference between the final calibration quantity determined by the calibration quantity control unit in the process of the cycle immediately before the certain cycle and the first calibration quantity generated by the first calibration quantity calculation unit.

3. a final calibration amount buffer unit that stores the final calibration amount determined by the calibration amount control unit as data; The likelihood input determination unit 2. The tracking device according to claim 1, wherein a process of determining the final posterior probability distribution is performed based on a ratio between the final calibration quantity determined by the calibration quantity control unit in the process one cycle before the certain cycle and the first calibration quantity generated by the first calibration quantity calculation unit.

4. a prior probability buffer unit that stores the prior probability distribution generated by the prior probability distribution generation unit as data, A tracking device as described in any one of claims 1 to 3, wherein the prior probability distribution generation unit stores the prior probability distribution for the period following the certain period, generated from the final posterior probability distribution determined by the likelihood input determination unit, in the prior probability buffer unit.

5. a storage device having a likelihood input determination threshold storage unit that stores a likelihood input determination threshold that is set in advance based on the continuity of the signal related to the tracking target, The tracking device according to any one of claims 1 to 4, wherein the likelihood input determination unit compares the difference between the final test amount in the processing one cycle before the certain cycle and the first test amount with the likelihood input determination threshold to determine the final posterior probability distribution.

6. 6. The tracking device according to claim 1, further comprising an orientation estimation unit that estimates the orientation of the tracking target from the tentative posterior probability distribution or the final posterior probability distribution.

7. a receiver array having a plurality of receivers and a signal converter that converts signals received by the receivers into digital data and generates observation value data; The tracking device according to any one of claims 1 to 6, A tracking system comprising:

8. A tracking method performed by a data processing device that periodically processes observation data obtained from a signal related to a tracking target and tracks the tracking target, comprising: The method performed by the data processing device includes: calculating likelihoods in a plurality of directions from the observation value data in a certain period and generating a likelihood function with the directions as parameter variables; generating a tentative posterior probability distribution from the likelihood function; generating a first statistic from the hypothetical posterior probability distribution; determining a final posterior probability distribution for the period; generating a prior probability distribution relating to the direction of the tracking target based on the final posterior probability distribution; generating a second statistic from the prior probability distribution; determining either the first assay amount or the second assay amount as a final assay amount based on the result of a comparison based on the first assay amount and the second assay amount; determining whether or not the tracking target exists in the certain period based on the final verification amount; and the step of determining the final posterior probability distribution for the certain cycle determines, based on the first validation quantity and the final validation quantity for the process one cycle before the certain cycle, either the tentative posterior probability distribution or the prior probability distribution for the process one cycle before the certain cycle, as the final posterior probability distribution for the certain cycle; Tracking method.

9. A tracking method program for tracking a tracking target based on observation value data obtained from a signal related to the tracking target, a step of calculating likelihoods in a plurality of directions from the observation value data in a certain period and generating a likelihood function with the directions as parameter variables; generating a tentative posterior probability distribution from the likelihood function; generating a first statistic from the provisional posterior probability distribution; performing a process for determining a final posterior probability distribution of the certain period; generating a priori probability distribution relating to the direction of the tracking target based on the final posterior probability distribution; generating a second estimator from the prior probability distribution; performing a process in which either the first assay amount or the second assay amount is set as a final assay amount based on a comparison result based on the first assay amount and the second assay amount; performing a process of determining whether or not the tracking target exists in the certain period based on the final verification amount; This is done by a computer. the step of performing processing to determine the final posterior probability distribution for the certain cycle includes determining, based on the first validation quantity and the final validation quantity in processing one cycle before the certain cycle, either the tentative posterior probability distribution or the prior probability distribution in processing one cycle before the certain cycle as the final posterior probability distribution for the certain cycle. Tracking method program.

Images