JP6299172B2 - Information processing apparatus, information processing method, and program - Google Patents

Description

  The present invention relates to an information processing apparatus, an information processing method, and a program.

  In recent years, techniques for determining the continuity of signals (sensor signals) obtained from sensors such as a vibration sensor, an acceleration sensor, and a Doppler sensor have been developed for watching a room and recognizing the state of a target object. The determination of continuity is performed by quantitatively evaluating the degree of temporal change in the signal cycle. By determining the continuity of the sensor signal, the nature of the signal is clarified, and the state of the target object and the target space observed by the sensor and the change in the state can be clarified.

  As such a technique, for example, in Patent Document 1 below, a zero cross value is applied to fluctuations in a very low frequency band of a biological signal, and a time series change of an average value of frequencies obtained for each time window is changed in frequency. A technique for obtaining a time series waveform is disclosed.

  Patent Document 2 below discloses a technique for collecting a living sound in a home with a sound sensor, extracting a component for each frequency by a short-time Fourier transform, and grasping a user's life pattern.

  Further, in the following non-patent document 1, in the time-varying coefficient autoregressive model expressed by the following mathematical formula 1, using data obtained for each time step, the current one to the previous N (N is a natural number) The technique which calculates | requires the autoregressive coefficient of this using a Kalman filter is disclosed.

JP2013-111103A JP2011-237865A

Genshiro Kitagawa. Introduction to time series analysis. Iwanami Shoten, 2005.

  However, the technique disclosed in the above document has a problem in terms of accuracy and calculation cost when determining the continuity of the signal output from the sensor in real time.

  For example, the zero-cross value used in the technique disclosed in Patent Document 1 has a large amplitude frequency, so the signal-to-noise ratio (S / N ratio) is low and the noise amplitude is relatively large. In some cases, it was difficult to detect the exact frequency of the desired signal.

  Since the technique disclosed in Patent Document 2 uses Fourier transform, the amount of calculation when performing calculations using data for each time window is large. For this reason, it is difficult for a terminal with low calculation capability to perform a real-time operation of sequentially extracting frequency components. In addition, when the number of data is small, there is a problem that sufficient frequency resolution cannot be obtained. Furthermore, when the desired signal and the noise frequency overlap in the frequency domain, it is difficult to separate the noise and the desired signal.

  In the technique disclosed in Non-Patent Document 1, when the autoregressive coefficient of a short time step is used, the calculation cost is small, but only local signal features are reflected in the autoregressive coefficient. It is difficult to accurately predict a stationary signal having a large number of samples per round. In addition, obtaining an autoregressive coefficient using observation values of a large number of samples is expensive in calculation and difficult to operate on a terminal having low calculation capability.

  Accordingly, the present invention has been made in view of the above problems, and an object of the present invention is to perform real-time prediction of a steady signal output from a sensor with a small amount of calculation with high accuracy. Another object of the present invention is to provide a new and improved information processing apparatus, information processing method and program.

  In order to solve the above problems, according to an aspect of the present invention, an extraction unit that extracts a period of a time-series signal output from a sensor, and an autoregressive coefficient corresponding to the period extracted by the extraction unit are provided. Provided is an information processing apparatus comprising: a coefficient calculation unit that sequentially calculates; and a prediction unit that predicts the time-series signal using an autoregressive model equation using the autoregressive coefficient calculated by the coefficient calculating unit. Is done.

  The extraction unit may sequentially extract the period.

  The extraction unit may extract a time step that takes at least one of a maximum value and a maximum value of a sequential autocovariance function as the period.

  The extraction unit may extract a time step taking at least one of a maximum value and a maximum value of a sequential autocorrelation function as the period.

  The extraction unit may extract, as the period, a time step that exceeds a threshold among time steps that take at least either the maximum value or the maximum value in the time-series signal in a predetermined section.

  The coefficient calculation unit may calculate the autoregressive coefficient each time the time series signal corresponding to a unit quantity time step is output.

  The coefficient calculation unit may calculate the autoregressive coefficient using a Kalman filter.

  The coefficient calculation unit may sequentially calculate an autoregressive coefficient at an arbitrary time step in addition to the autoregressive coefficient corresponding to the period.

  The information processing apparatus further includes a determination unit that determines the continuity of the time series signal based on a comparison result between the predicted value of the time series signal predicted by the prediction unit and the observation value output from the sensor. You may prepare.

  The determination unit may determine that the difference between the predicted value and the observed value is greater than or equal to a threshold value, and may determine that the difference is less than the threshold value.

  The sensor may be at least one of a Doppler sensor, a vibration sensor, and an acceleration sensor.

  In order to solve the above problem, according to another aspect of the present invention, a step of extracting a period of a time-series signal output from a sensor and an autoregressive coefficient corresponding to the extracted period are sequentially obtained. There is provided an information processing method comprising the steps of: calculating the time series signal by using an equation of an autoregressive model using the calculated autoregressive coefficient.

  In order to solve the above problem, according to another aspect of the present invention, a computer is provided with an extraction unit that extracts a period of a time-series signal output from a sensor, and the period extracted by the extraction unit. A coefficient calculation unit that sequentially calculates a corresponding autoregressive coefficient, and a prediction unit that performs prediction of the time series signal using an equation of an autoregressive model using the autoregressive coefficient calculated by the coefficient calculation unit, and A program for functioning is provided.

  As described above, according to the present invention, real-time prediction of a steady signal output from a sensor can be accurately performed with a smaller amount of calculation.

It is a block diagram which shows an example of a structure of the stationarity determination apparatus which concerns on one Embodiment of this invention. It is a flowchart which shows operation | movement of the stationarity determination apparatus which concerns on one Embodiment of this invention. It is a figure which shows the prediction error by the stationarity determination apparatus which concerns on a comparative example. It is a figure which shows the prediction error by the stationarity determination apparatus which concerns on one Embodiment of this invention.

  Exemplary embodiments of the present invention will be described below in detail with reference to the accompanying drawings. In addition, in this specification and drawing, about the component which has the substantially same function structure, duplication description is abbreviate | omitted by attaching | subjecting the same code | symbol.

<1. Overview of stationarity determination device>
First, an outline of a continuity determination device according to an embodiment of the present invention will be described. The stationarity determination device according to the present embodiment determines the stationarity of the period of the sensor signal in real time. The stationarity determination device according to this embodiment finds a change point about the motion of the target object and the state of space by analyzing the stationarity of sensor signals obtained from vibration sensors, acceleration sensors, Doppler sensors, and the like. can do. As described above, the technique using the zero cross value, the Fourier transform, and the time-varying coefficient autoregressive model described in the above document has problems in terms of accuracy and calculation cost.

  Therefore, the stationarity determination device according to the present embodiment can improve the technique disclosed in Non-Patent Document 1 and determine the stationarity of a sensor signal in real time while accurately reducing the amount of calculation. I made it. Specifically, the stationarity determination device obtains a cycle of the sensor signal by a sequential autocovariance function, and calculates a predicted value by an autoregressive model configured by terms of time steps corresponding to the cycle. The stationarity determination device can obtain a predicted value based on an autoregressive model of a stationary signal with a large number of samples per cycle by focusing on the time step corresponding to the cycle with high accuracy and low calculation cost. . Here, a large number of samples per cycle means the number of samples that is difficult to operate in real time due to the problem of calculation cost with the technique described in the above document, and depends on the sample interval and the length of the cycle. Value. If the sample interval is short, the number of samples is large even in a short time, and if the sample interval is long, the number of samples can be small even in a long time. The stationarity determination device determines the stationarity of the sensor signal using the predicted value.

  The process by the continuity determination device will be described in more detail. The stationarity determination device extracts a signal period from a time-series signal output from a sensor using a sequential autocovariance function with a relatively small amount of calculation. Next, the stationarity determination apparatus sequentially calculates autoregressive coefficients using autoregressive coefficients of time steps corresponding to the extracted periods. The stationarity determination device quantitatively determines stationarity using the difference between the predicted value and the actual value of the autoregressive model configured by the calculated autoregressive coefficient as a prediction error, and using the prediction error as a signal fluctuation. .

  In the present embodiment, even when noise having the same frequency as that of the desired signal exists, the continuity of the desired signal can be determined by evaluating the fluctuation of the period. In addition, by using the autoregressive coefficient of the time step corresponding to the extracted period, it is possible to predict a steady signal having a large number of samples per period with high accuracy and a small amount of calculation. As a result, in this embodiment, the accuracy of distinguishing between stationary signals and non-stationary signals is improved.

  The overview of the stationarity determination device has been described above. Next, a configuration example of the continuity determination device will be described with reference to FIG.

<2. Configuration example of stationarity determination device>
FIG. 1 is a block diagram illustrating an example of a configuration of a continuity determination device according to an embodiment of the present invention. As shown in FIG. 1, the stationarity determination device 1 includes a sensor 10, an AD (Analog-to-Digital) converter 20, a signal period extraction unit 30, a stationarity determination unit 40, and an output unit 50.

(Sensor 10)
The sensor 10 is an apparatus that observes an arbitrary target space or target object and outputs an observed time series signal (sensor signal). The sensor 10 is realized as a vibration sensor, an acceleration sensor, or a Doppler sensor, for example. In the present embodiment, it is assumed that the sensor signal is an analog signal.

(AD converter 20)
The AD converter 20 is an electronic circuit that converts an analog electric signal into a digital electric signal. The AD converter 20 converts the sensor signal output from the sensor 10 from an analog signal to a digital signal and outputs the signal.

(Signal period extraction unit 30)
The signal period extraction unit 30 has a function as an extraction unit that extracts the period of the time-series signal output from the sensor 10. The signal period extraction unit 30 sequentially extracts the period of the time-series signal in order to ensure real-time continuity determination accuracy by the continuity determination unit 40. Various means can be conceived for the signal period extraction unit 30 to sequentially extract the periods. For example, the signal period extraction unit 30 extracts a time step that takes a peak of a sequential autocovariance function or a sequential autocorrelation function as a period. The peak here is at least one of a maximum value and a maximum value. The maximum value represents the main period of the signal. The maximum value represents the period of each signal when a plurality of signals having different periods are mixed. The time interval between each time step of the sensor signal and the current time is also referred to as lag time.

  The sequential covariance function and the sequential autocorrelation function are suitable as means for extracting a period in real time because the amount of calculation is small and the frequency resolution can be finely taken. Since the sequential autocovariance function is obtained by multiplying correlation by variance, the signal period extraction unit 30 extracts the period by removing the influence of noise having a small amplitude when the sequential autocovariance function is used. Can do. On the other hand, when the sequential autocorrelation function is used, the signal period extraction unit 30 can extract the period even for a minute signal. The signal period extraction unit 30 selects whether to use a sequential autocovariance function or a sequential autocorrelation function based on the type of sensor 10, the magnitude of noise, and the like.

  The signal period extraction unit 30 extracts one or a plurality of peaks. Here, in consideration of the prediction accuracy by the prediction unit 43, the signal period extraction unit 30 extracts all the peaks of the time-series signal, and the prediction unit 43 has a time step term corresponding to the peak. It is desirable to predict with. However, from the viewpoint of reducing the calculation cost, the signal period extraction unit 30 extracts, as a period, a time step in which the signal value exceeds the threshold among time steps that take at least one of the maximum value and the maximum value. Further, from the viewpoint of reducing the calculation cost, the signal period extraction unit 30 may extract a peak from a time-series signal in a predetermined section by dividing the period.

  In addition to the method using the sequential autocovariance function or the sequential autocorrelation function, the signal period extraction unit 30 may extract a time step corresponding to a known periodicity for the time series signal as a period. The known periodicity is a periodicity extracted by, for example, Fourier transform or zero cross value.

(Stability determination unit 40)
The stationarity determination unit 40 has a function of determining the stationarity of the time-series signal output from the sensor 10 using the period extracted by the signal period extraction unit 30. As illustrated in FIG. 1, the continuity determination unit 40 functions as a coefficient calculation unit 41, a prediction unit 43, and a determination unit 45.

Coefficient calculation unit 41
The coefficient calculation unit 41 has a function of sequentially calculating autoregressive coefficients of time steps corresponding to the period extracted by the signal period extraction unit 30. Specifically, the coefficient calculation unit 41 calculates an autoregressive coefficient each time a time-series signal for a unit quantity of time steps is output from the sensor 10. The unit quantity time step may be, for example, one step or a time step of a predetermined number of seconds.

  There are various means by which the coefficient calculation unit 41 sequentially extracts autoregressive coefficients. For example, the coefficient calculation unit 41 calculates an autoregressive coefficient using a Kalman filter. As a means for sequentially calculating autoregressive coefficients other than the Kalman filter, for example, a SDAR (Sequentially Distinguishing AR model Estimating) algorithm can be considered. However, in the SDAR algorithm, since it is necessary to calculate the autoregressive coefficients of all time steps, it is possible to selectively calculate only the autoregressive coefficients of the time steps corresponding to the periods extracted by the signal period extracting unit 30. Can not. Therefore, the coefficient calculation unit 41 can calculate the autoregressive coefficient at a lower calculation cost when the Kalman filter is used than when the SDAR algorithm is used. In addition, the coefficient calculation unit 41 may calculate the autoregressive coefficient using, for example, a particle filter.

  The autoregressive coefficient calculated by the coefficient calculating unit 41 is an autoregressive coefficient of a time step corresponding to the period extracted by the signal period extracting unit 30, and constitutes the following self model used for prediction by the prediction unit 43 described later. . For this reason, since the autoregressive model expresses the periodicity of the time-series signal well, the prediction unit 43 can accurately predict a steady signal having a large number of samples per cycle.

  The coefficient calculation unit 41 may sequentially calculate an autoregressive coefficient at an arbitrary time step in addition to the autoregressive coefficient at a time step corresponding to the period extracted by the signal period extraction unit 30. For example, the coefficient calculation unit 41 calculates autoregressive coefficients for several steps nearest to the current time. Thereby, the prediction part 43 can perform the prediction which considered the observation value about the last several steps. In consideration of the prediction accuracy by the prediction unit 43, the signal period extraction unit 30 extracts all the peaks of the time-series signal, and the prediction unit 43 predicts with an autoregressive model having a time step term corresponding to the peak. Is desirable. However, from the viewpoint of reducing the calculation cost, it is conceivable that the signal period extraction unit 30 extracts peaks by dividing the period. In such a case, the prediction unit 43 can compensate for a decrease in prediction accuracy due to the division of the period by performing the prediction in consideration of the observation values for the most recent steps.

・ Prediction unit 43
The prediction unit 43 has a function of predicting a time series signal using an autoregressive model using the autoregressive coefficient calculated by the coefficient calculating unit 41. The prediction unit 43 predicts a time-series signal using Equation 13 described later. Referring to Equation 1 and Equation 13, the prediction unit 43 predicts a time-series signal using an autoregressive model (time-varying coefficient autoregressive model) as in the technique described in Non-Patent Document 1. However, in the said nonpatent literature 1, as shown to the said Numerical formula 1, it is necessary to calculate about all the time steps of an object period. On the other hand, the prediction unit 43 only has to calculate only the time step of the period extracted by the signal period extraction unit 30 as shown in Equation 13. For this reason, in this embodiment, compared with the technique described in the said nonpatent literature 1, calculation amount can be reduced. The prediction unit 43 outputs the prediction result to the determination unit 45.

・ Determining unit 45
The determination unit 45 determines the continuity of the time-series signal based on the comparison result between the predicted value of the time-series signal predicted by the prediction unit 43 and the actual value (observed value) output from the sensor 10. It has a function. More specifically, the determination unit 45 determines that the difference between the predicted value and the actual value is not less than the threshold value, and determines that the difference is less than the threshold value. The difference between the predicted value and the actual value is hereinafter also referred to as a prediction error. The determination unit 45 outputs the determination result to the output unit 50.

(Output unit 50)
The output unit 50 outputs the determination result by the continuity determination unit 40 by video, image, audio, or the like. The output unit 50 is realized by, for example, a CRT (Cathode Ray Tube) display device, a liquid crystal display (Liquid Crystal Display) device, a speaker, or the like.

  The configuration example of the continuity determination device according to the present embodiment has been described above.

<3. Operation of Stationaryness Determination Device>
In this section, the operation for determining the presence or absence of a person in an environment where there is an object that moves mechanically and periodically, such as a fan, will be described. The sensor signal when a person is present is an unsteady signal, and the sensor signal when no person is present is a steady signal. The operation of the stationarity determination device 1 includes first acquisition of a signal from the sensor 10 and conversion thereof, second extraction of a signal period, third prediction using an autoregressive coefficient and determination of stationarity based on a prediction error. There are three stages. Hereinafter, the operation of each stage will be described with reference to FIGS. FIG. 2 is a flowchart showing the operation of the stationarity determination device 1 according to an embodiment of the present invention.

[3-1. Signal acquisition from sensor and its conversion]
As shown in FIG. 2, first, in step S102, the stationarity determination device 1 acquires a sensor signal. Specifically, the sensor 10 outputs the sensing result as an analog time series signal.

  Next, in step S104, the stationarity determination device 1 converts the sensor signal into a digital signal. Specifically, the AD converter 20 converts an analog time series signal output from the sensor 10 into a digital signal. The analog signal acquired by the sensor 10 is generally a minute signal. For this reason, in order to suppress the influence of noise and improve the S / N ratio, the stationarity determination apparatus 1 may amplify with an amplifier or perform frequency filtering with an analog filter or a digital filter.

[3-2. Extraction of signal period]
Next, in step S106, the stationarity determination apparatus 1 obtains a sequential autocovariance function at each lag time. Specifically, the signal period extraction unit 30 sequentially obtains an autocovariance function based on the sensor signal output from the AD converter 20. The sequential autocovariance function at a lag time τ at a certain time n is expressed by the following Equation 3. More specifically, the signal period extraction unit 30 sequentially determines the observed value y _n (n = 1, 2,... N) at the sequential average u _{n at} a certain time n and the lag time τ at a certain time n. The autocovariance function C (n, τ) is calculated by the recurrence formulas of the following formulas 2 and 3, respectively.

Here, r is a forgetting factor and takes a value in the range of 0 <r <1. Different forgetting factors may be used in Equation 2 and Equation 3. The signal period extraction unit 30 obtains a sequential autocovariance function in the time range, where 0 ≦ τ ≦ τ _max is a period in which the signal period exists.

  Next, in step S108, the stationarity determination apparatus 1 extracts one or a plurality of raglimes whose sequential autocovariance function has a maximum value or a maximum value. Specifically, the signal period extraction unit 30 extracts one or a plurality of lag times τ at which the sequential autocovariance function expressed by Equation 3 has a maximum value or a lag time τ at which the maximum value is reached.

[3-3. Prediction using autoregressive coefficient and determination of stationarity by prediction error]
Next, in step S110, the stationarity determination device 1 configures a time-varying coefficient autoregressive model corresponding to the extracted lag time. Specifically, first, the coefficient calculation unit 41 sequentially calculates the autoregressive coefficient of the time step corresponding to the period extracted by the signal period extraction unit 30 in step S108. The coefficient calculation unit 41 calculates autoregressive coefficients using a Kalman filter as described below.

  The Kalman filter is modeled by a state equation represented by Equation 4 and an observation equation represented by Equation 5, which give a relationship between a state vector x composed of m-dimensional unknown parameters and a scalar observation value y.

Here, u _n is m-dimensional system noise, v _n is the one-dimensional observation noise, the following equation 6, as in Equation 7, the mean 0, variance-covariance matrix follows Q, the normal distribution of the R, respectively Assume that

Here, I is a unit matrix. K _n is called the Kalman gain is expressed by the following equation 12.

  When obtaining the autoregressive coefficient using the Yule-Walker equation, it is necessary to obtain the lag step in order from 1. However, as shown in the above equation 13, if the Kalman filter is used, only the necessary ordinal coefficient can be obtained. it can. That is, the coefficient calculating unit 41 can calculate the autoregressive coefficient at a lower calculation cost by using only the autoregressive coefficient corresponding to the lag time extracted by the signal period extracting unit 30 as a calculation target. As described above, in step S110, the coefficient calculation unit 41 configures the autoregressive model shown in Equation 13 as a time-varying coefficient autoregressive model corresponding to the lag time extracted by the signal period extraction unit 30. .

In step S112, the stationarity determination device 1 obtains a prediction error. Specifically, first, the prediction unit 43 obtains a predicted value based on the autoregressive model shown in the mathematical formula 13. Then, the determination unit 45, the difference between the actual value output by the prediction value and the sensor 10 predicted by the prediction unit 43 is obtained as the prediction error e _n represented by the following equation 14.

  Here, referring to FIG. 3 and FIG. 4, an autoregressive model (comparative example) composed of autoregressive coefficients from one step before to N steps before the current time, and an autoregressive coefficient corresponding to the period of the signal The difference in prediction error from the autoregressive model (this embodiment) configured by FIG. 3 is a diagram illustrating a prediction error by the continuity determination device according to the comparative example. FIG. 4 is a diagram showing a prediction error by the stationarity determination device 1 according to an embodiment of the present invention.

  In FIG. 3, as a comparative example, the order is 3, and prediction errors based on the autoregressive model for the three most recent time steps d1, d2, and d3 are shown. As shown in FIG. 3, since the most recent observation value in the past is used for prediction, the autoregressive model cannot express the period exceeding the three time steps of interest, and the prediction value is the latest observation value. The prediction error may increase. In particular, if the signal is locally irregular within or around the three time steps of interest, the prediction error will be large.

  FIG. 4 shows the prediction error by the autoregressive model for the most recent two time steps d2 and d3 and the time step d1 corresponding to the period extracted by the sequential autocovariance function, where the order is 3. As shown in FIG. 4, by adopting a term corresponding to the period extracted by the sequential autocovariance function in the autoregressive model, the autoregressive model can express the periodicity of the signal. The prediction error can be reduced as compared with the comparative example shown. This autoregressive model does not depend excessively on the most recent observations, so the prediction error increases even if the signal is locally irregular within or around the three time steps of interest. Can be suppressed.

  The processing in step S112 has been described above. Next, in step S114, the stationarity determination device 1 determines whether the prediction error is greater than or equal to a threshold value. Specifically, the determination unit 45 determines whether or not the prediction error expressed by Equation 14 is greater than or equal to a threshold value.

  If it is determined that the prediction error is greater than or equal to (S114 / YES), in step S116, the stationarity determination unit 40 determines that the sensor signal is an unsteady signal. For example, when a person is present in the sensing target space of the sensor 10, a signal resulting from the person's breathing or movement is observed, so the continuity determination unit 40 determines that the sensor signal is an unsteady signal.

  On the other hand, when it is determined that the prediction error is less (S114 / NO), in step S118, the stationarity determination unit 40 determines that the sensor signal is a stationar signal. For example, when there is no person in the sensing target space of the sensor 10, only the stationary signal due to the periodic movement of the electric fan is observed, so the continuity determination unit 40 determines that the sensor signal is a stationary signal. To do.

  In step S120, the stationarity determination device 1 outputs a determination result. Specifically, the output unit 50 outputs the determination result determined by the determination unit 45 in step S116 or S118.

Next, in step S122, the stationarity determination device 1 determines whether or not to end the process.
When it is determined not to end (S122 / NO), the process returns to step S102 again, and the stationarity determination device 1 repeats the stationarity determination. Thereby, the stationarity determination apparatus 1 can determine the stationarity of the sensor signal in real time. On the other hand, when it determines with complete | finishing (S122 / YES), the stationarity determination apparatus 1 complete | finishes a process.

  The operation of the continuity determination device 1 according to this embodiment has been described above.

(Supplement)
As described above, the signal period extraction unit 30 can extract the period using the sequential autocorrelation function in addition to the sequential autocovariance function. In this case, the signal period extraction unit 30 replaces the equation 3 above with the sequential autocovariance function C (n, τ) at time n shown in the equation 3 divided by the sequential variance v (n). The sequential autocorrelation function Rxx (n, τ) shown in FIG.

  Here, the sequential variance v (n) is given by a recurrence formula shown by the following formula 16.

  Here, r is a forgetting factor and takes a value in the range of 0 <r <1.

<4. Summary>
As described above, according to the present embodiment, it is possible to determine the continuity of a sensor signal with high accuracy and in real time with a smaller amount of calculation. More specifically, the stationarity determination device 1 uses the sequential autocovariance function or the sequential autocorrelation function to extract the period of the sensor signal with a smaller amount of calculation compared to the case where Fourier transform or the like is used. Can do. In addition, the stationarity determination apparatus 1 can reduce the prediction error of a periodic signal by constructing an autoregressive model with respect to a time step corresponding to the extracted period, so that the steady state of the sensor signal can be improved more accurately. Gender can be determined.

  The stationarity determination device 1 determines the stationarity of the sensor signal, for example, based on the output result from the Doppler sensor, the movement of a device that performs a mechanical operation at a constant period, such as an electric fan, It is possible to distinguish from such irregular movement with a large fluctuation of the period. The stationarity determination device 1 can also be applied to a human sensor that detects the presence or absence of a person with high accuracy.

  The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to such examples. It is obvious that a person having ordinary knowledge in the technical field to which the present invention pertains can come up with various changes or modifications within the scope of the technical idea described in the claims. Of course, it is understood that these also belong to the technical scope of the present invention.

  For example, in the above embodiment, the sensor 10, the AD converter 20, the signal period extraction unit 30, the continuity determination unit 40, and the output unit 50 are integrated, but the present invention is not limited to such an example. For example, each component may be configured separately. Specifically, a PC (Personal Computer), a mobile terminal, a server on the network, and the like function as the signal period extraction unit 30 and the continuity determination unit 40, and the determination result is obtained based on the output from the sensor 10 and the AD converter 20. You may output to the output part 50.

  Further, it is possible to create a computer program for causing hardware such as a CPU, a ROM, and a RAM built in the information processing apparatus to perform the same functions as the components of the continuity determination apparatus 1. A recording medium recording the computer program is also provided.

DESCRIPTION OF SYMBOLS 1 Steadyness determination apparatus 10 Sensor 20 AD converter 30 Signal period extraction part 40 Steadyness determination part 41 Coefficient calculation part 43 Prediction part 45 Determination part 50 Output part

Claims (13)

An extraction unit that sequentially extracts the period of the time-series signal output from the sensor;
A coefficient calculating unit that sequentially calculates autoregressive coefficients corresponding to the period extracted by the extracting unit;
A prediction unit that predicts the time-series signal by an equation of an autoregressive model using the autoregressive coefficient calculated by the coefficient calculating unit;
Equipped with a,
The information processing apparatus , wherein the extraction unit extracts a time step taking at least one of a maximum value and a maximum value of a sequential autocovariance function as the period . An extraction unit that sequentially extracts the period of the time-series signal output from the sensor;
A coefficient calculating unit that sequentially calculates autoregressive coefficients corresponding to the period extracted by the extracting unit;
A prediction unit that predicts the time-series signal by an equation of an autoregressive model using the autoregressive coefficient calculated by the coefficient calculating unit;
Equipped with a,
The information processing apparatus , wherein the extraction unit extracts, as the period, a time step that takes at least one of a maximum value and a maximum value of a sequential autocorrelation function . The extraction section, among the time step taking any at least of the maximum value or the maximum in the time-series signal of a predetermined period, the time step that exceeds the threshold value is extracted as the period, according to claim 1 or 2 Information processing device. The coefficient calculation unit, each time the time-series signal time step portion of the unit quantity is output, calculates the autoregressive coefficients, the information processing apparatus according to any one of claims 1-3. The information processing apparatus according to claim 4 , wherein the coefficient calculation unit calculates the autoregressive coefficient using a Kalman filter. The coefficient calculation unit, in addition to the autoregressive coefficients corresponding to the period to sequentially calculate the autoregressive coefficients of the arbitrary time step, the information processing apparatus according to any one of claims 1-5. The information processing apparatus further includes a determination unit that determines the continuity of the time series signal based on a comparison result between the predicted value of the time series signal predicted by the prediction unit and the observation value output from the sensor. The information processing apparatus according to any one of claims 1 to 6 , further comprising: The determination unit, the difference between the predicted value and the observed value is determined to be non-stationary if it is more than the threshold value, it determines that the constant in the case is less than the threshold value, according to claim 7 Information processing device. The sensor Doppler sensor, the vibration sensor is at least one of an acceleration sensor, the information processing apparatus according to any one of claims 1-8. Sequentially extracting the period of the time-series signal output from the sensor;
Sequentially calculating autoregressive coefficients corresponding to the extracted periods;
Predicting the time-series signal according to an equation of an autoregressive model using the calculated autoregressive coefficient;
Equipped with a,
The information extracting method, wherein the extracting step extracts a time step taking at least one of a maximum value and a maximum value of a sequential autocovariance function as the period . Sequentially extracting the period of the time-series signal output from the sensor;
Sequentially calculating autoregressive coefficients corresponding to the extracted periods;
Predicting the time-series signal according to an equation of an autoregressive model using the calculated autoregressive coefficient;
Equipped with a,
The information extracting method wherein the extracting step extracts a time step taking at least one of a maximum value and a maximum value of a sequential autocorrelation function as the period . Computer
An extraction unit that sequentially extracts the period of the time-series signal output from the sensor;
A coefficient calculating unit that sequentially calculates autoregressive coefficients corresponding to the period extracted by the extracting unit;
A prediction unit that predicts the time-series signal by an equation of an autoregressive model using the autoregressive coefficient calculated by the coefficient calculating unit;
To function as,
The said extraction part is a program which extracts the time step which takes at least one of the maximum value or maximum value of a sequential autocovariance function as said period . Computer
An extraction unit that sequentially extracts the period of the time-series signal output from the sensor;
A coefficient calculating unit that sequentially calculates autoregressive coefficients corresponding to the period extracted by the extracting unit;
A prediction unit that predicts the time-series signal by an equation of an autoregressive model using the autoregressive coefficient calculated by the coefficient calculating unit;
To function as,
The said extraction part is a program which extracts the time step which takes at least one of the maximum value or maximum value of a sequential autocorrelation function as said period .

Images