JP6516042B2 - Signal processing device, detection device, sensor, electronic device and moving body - Google Patents

Description

  The present invention relates to a signal processing device, a detection device, a sensor, an electronic device, a moving object, and the like.

  BACKGROUND ART Gyro sensors for detecting physical quantities that change due to external factors are incorporated in electronic devices such as digital cameras and smartphones, and mobile bodies such as cars and airplanes. Such a gyro sensor detects a physical quantity such as angular velocity, and is used for so-called image stabilization, attitude control, GPS autonomous navigation, and the like.

  The gyro sensor outputs the detected physical quantity such as angular velocity as a detection voltage signal. The application that has received the detected voltage signal integrates the angular velocity and the like obtained by the detected voltage signal by software processing to determine the angle, the velocity, the distance, and the like.

Unexamined-Japanese-Patent No. 2004-347505 JP 2002-236302 A

  Now, in the detection voltage signal from the physical quantity transducer, there is a DC offset due to, for example, the temperature characteristic of the physical quantity transducer. There is a problem that the DC offset causes an error when obtaining desired information from the detection voltage signal. For example, in the above-described gyro sensor, since the detection voltage signal including the DC offset is integrated when the angle or the like is obtained, the error becomes very large.

  As a method of removing a DC offset, Patent Document 1 discloses a method of processing a detection voltage signal with a high pass filter. Further, Patent Document 2 discloses a method of adding a bias of a predetermined voltage to a detection voltage signal. However, in the method of Patent Document 1, since the characteristic of the high-pass filter is fixed, there is a problem that the characteristic affects the transient response of the detection voltage signal. The method of Patent Document 2 has a problem that it is difficult to follow changes in DC offset.

  According to some aspects of the present invention, it is possible to provide a signal processing device, a detection device, a sensor, an electronic device, a moving body, and the like capable of extracting a DC component with improved transient response and followability.

  The present invention has been made to solve at least a part of the above-mentioned problems, and can be realized as the following modes or embodiments.

  One aspect of the present invention is a noise estimation unit that estimates observation noise and system noise that dynamically change according to an input signal, and Kalman filtering based on the observation noise and the system noise to obtain the input signal. And a Kalman filter for extracting a DC component of

  According to one aspect of the present invention, the noise estimation unit dynamically changes observation noise and system noise in accordance with an input signal and supplies it to the Kalman filter, and the Kalman filter dynamically changes the observation noise and system. Performs Kalman filter processing in response to noise. By supplying observation noise and system noise from the outside of the Kalman filter in this way, it is possible to control the characteristics of the Kalman filter, and it becomes possible to extract a DC component with improved transient response and followability.

  In one aspect of the present invention, the noise estimation unit may estimate the system noise by estimating the observation noise based on the input signal and performing gain processing on the estimated observation noise. .

  In this way, observation noise can be dynamically changed according to the input signal, and system noise can be adjusted by gain processing on the observation noise. This makes it possible to control the Kalman filter to have a desired characteristic.

  In one aspect of the present invention, the noise estimation unit may estimate the observation noise by performing a mean square of the input signal and performing a limiter process on the signal after the mean square.

  Observation noise does not become zero in real observation. In this respect, according to one aspect of the present invention, even if the observation noise estimated from the input signal becomes zero, the lower limit value of the observation noise can be limited by the limiter processing, so non-zero observation noise is output to the Kalman filter. it can.

  In one aspect of the present invention, the noise estimation unit performs the gain processing on the observation noise with a gain set based on a target cutoff frequency of a low pass filter operation in a convergence state of the Kalman filter. It is also good.

  When sufficient time has passed since the Kalman filter started to operate, the Kalman filter converges on the filter characteristics including the low pass filter characteristics. In this convergence state, the cutoff frequency of the low pass filter characteristic is determined by the gain of gain processing. That is, by setting the gain based on the target cutoff frequency, it is possible to obtain low-pass filter characteristics of a desired cutoff frequency in the convergence state.

  In one aspect of the present invention, the noise estimation unit may perform high-pass filter processing on the input signal and may perform processing to increase the observation noise based on the signal after the high-pass filter processing.

  If the input signal is not DC but fluctuates, the estimation accuracy of the DC component of the Kalman filter is reduced. In this respect, according to one aspect of the present invention, by increasing the observation noise based on the input signal after high-pass filter processing, the Kalman gain becomes smaller, and the degree of confidence of the Kalman filter decreases. it can. Thereby, the estimation accuracy of the DC component can be improved.

  In one aspect of the present invention, the noise estimation unit may perform peak hold processing on the signal after the high-pass filter processing, and increase the observation noise based on the signal after the peak hold processing.

  When the input signal is fluctuating, the signal after high-pass filtering may be zero. When the signal is zero, the observation noise does not increase, so the reliability of the observation value increases and the estimation accuracy of the DC component decreases. In this respect, according to one aspect of the present embodiment, it is possible to suppress that the signal becomes zero by the peak hold processing, so observation noise is increased while the input signal is fluctuating, and the estimation accuracy of the DC component is improved. it can.

  Another aspect of the present invention is a drive circuit for driving a physical quantity transducer, a detection circuit for receiving a detection signal from the physical quantity transducer and detecting a physical quantity signal according to the physical quantity, and the physical quantity signal as the input signal And a signal processing apparatus according to any of the above, which extracts the DC component as

  Yet another aspect of the present invention relates to a sensor comprising the detection device described above and the physical quantity transducer.

  Yet another aspect of the present invention relates to an electronic device including the signal processing device described in any of the above.

  Yet another aspect of the present invention relates to a mobile including the signal processing device according to any of the above.

The example of basic composition of electronic equipment. The structural example of a signal processing apparatus. The 1st example of detailed composition of a signal processing device. The 2nd detailed example of composition of a signal processing device. Explanatory drawing about a peak hold process. The 3rd detailed example of composition of a signal processing device. Explanatory drawing about DC designated value, deviation, and the maximum deviation value. The 4th detailed example of composition of a signal processing device. Explanatory drawing about the temperature characteristic of DC offset of a gyro sensor. Explanatory drawing about the time change of DC offset by temperature change. Configuration example of a gyro sensor and an electronic device. The structural example of a detection apparatus. Configuration example of a mobile.

  Hereinafter, preferred embodiments of the present invention will be described in detail. The present embodiment described below does not unduly limit the contents of the present invention described in the claims, and all the configurations described in the present embodiment are essential as the solution means of the present invention. Not necessarily.

1. Signal Processing Device Hereinafter, the signal processing device of the present embodiment will be described by taking the case of extracting a DC offset from the physical quantity signal of the physical quantity transducer as an example. Note that the present embodiment is applicable not only to the physical quantity signal of the physical quantity transducer, but also to any apparatus for extracting a DC component from the input signal.

  FIG. 1 shows an example of the basic configuration of an electronic device to which the signal processing apparatus 100 of the present embodiment is applied. The electronic device includes the physical quantity transducer 15, the detection circuit 60, the signal processing device 100, and the processing unit 520.

  The physical quantity transducer 15 converts physical quantities such as angular velocity, acceleration, angular acceleration, or velocity into an electrical signal (voltage, current). As the physical quantity transducer 15, various devices (sensors) such as a vibrator in a gyro sensor or an acceleration sensor can be assumed. The detection circuit 60 performs detection processing (synchronous detection, etc.) of a physical quantity such as angular velocity, acceleration, angular acceleration or velocity based on the detection signal TQ (current signal, voltage signal, etc.) from the physical quantity transducer 15. A physical quantity signal PI (detection data) corresponding to the physical quantity is output to the signal processing apparatus 100.

  The signal processing apparatus 100 extracts a DC component DCQ from the physical quantity signal PI, subtracts the DC component DCQ from the physical quantity signal PI, and outputs the signal after subtraction as an output signal PQ (output data). In FIG. 1, the subtraction unit is described outside the signal processing device 100 for convenience. The DC component DCQ is a component whose frequency is lower than a desired signal component desired to be extracted from the physical quantity signal PI. For example, when the physical quantity transducer 15 is a gyro sensor, the physical quantity signal PI includes an offset, and a change based on the offset becomes an actual signal component. The frequency of this signal component corresponds to the frequency of movement detected by the gyro sensor. The offset is not zero frequency because it varies with time due to temperature change etc., but it is a low frequency compared to the motion frequency.

  The processing unit 520 may use the output signal PQ from the detection device 20 as it is, but may integrate the output signal PQ and use the integration result value. For example, when the physical quantity transducer 15 is a gyro sensor, the detection device 20 outputs angular velocity data to the processing unit 520 based on the signal from the vibrator. Then, the processing unit 520 performs integration processing (integration processing) of angular velocity data to obtain angle data. Since the DC component is removed by the signal processing apparatus 100, accurate angle data with a small error can be obtained even if integration processing is performed.

  Now, as a first comparative example of the present embodiment, a method of removing a DC component using the above-described high pass filter can be considered. However, when the cutoff is set to a very low frequency in order to remove the DC component, there is a problem that it takes time for the high pass filter to converge. In order to improve the convergence time, the cutoff must be set to a certain frequency, which causes a problem that the pass band of the signal is limited. For example, with a gyro sensor, it is difficult to detect slow movement.

  As a second comparative example, a method may be considered in which an error due to a DC component is converged using an adaptive filter. For example, based on information from an inertial sensor other than the gyro sensor, an error such as an angle measured by the gyro sensor is converged. However, although this method adaptively changes the filter characteristics, there is a problem that another inertial sensor is required.

  FIG. 2 shows an example of the configuration of the signal processing apparatus 100 according to the present embodiment that can solve the problems as described above. The outline will be described below, and the detailed configuration and operation will be described later. Furthermore, although the case where the physical quantity transducer 15 is a gyro sensor will be described as an example below, as described above, the physical quantity transducer 15 is not limited to a gyro sensor.

  The signal processing apparatus 100 shown in FIG. 2 includes a noise estimation unit 110, a Kalman filter 120, and a subtraction unit 121.

The noise estimation unit 110 receives the physical quantity signal PI as an input signal (input data), and estimates observation noise σ _meas and system noise σ _sys dynamically changing according to the input signal PI. Specifically, the noise estimation unit 110 generates observation noise σ _meas and system noise σ _sys from the input signal PI, and changes observation noise σ _meas and system noise σ _sys according to the value of the input signal PI or its change. Change.

The Kalman filter 120 performs Kalman filtering based on the observation noise σ _meas and the system noise σ _sys to extract a DC component DCQ of the input signal PI. The subtraction unit 121 subtracts the DC component DCQ from the input signal PI, and outputs an output signal PQ. Kalman filter processing is processing that estimates the optimal state of the system using observed values acquired from the past to the present, assuming that noise (error) is included in the observation value and the variable representing the state of the system. . In the case of this embodiment, the observed value is the physical quantity signal PI of the gyro sensor, and the variable to be estimated is the DC component DCQ. Specifically, the state is estimated by repeatedly performing observation update (observation process) and time update (prediction process). The observation update is a process of updating the Kalman gain, the estimated value, and the error covariance using the observation value and the time update result. The time update is a process of predicting an estimated value at the next time and an error covariance using the result of the observation update.

  In a general Kalman filter, initial values of error covariance and system noise are given in advance as known ones. The error covariance is updated by observation update and time update. As described above, in the general Kalman filter, observation noise and system noise are not newly given from the outside in the middle of repeated updating.

On the other hand, in the present embodiment, the observation noise σ _meas and the system noise σ _sys are dynamically changed and supplied to the Kalman filter 120 from the outside. As described later in Equations (6) to (10) below, the observation noise σ _meas and the system noise σ _sys affect internal variables such as the Kalman gain g (k). That is, it means that the filter characteristics of the Kalman filter 120 can be controlled by controlling the observation noise σ _meas and the system noise σ _sys . In this embodiment, by using this, when the DC component of the physical quantity signal PI of the gyro sensor is not changed, the pass band can be set to a low frequency, and the pass band of the signal component can be extended to the low frequency side. . In addition, when the DC component changes, the observation noise σ _meas and the system noise σ _sys can be changed to widen the pass band and follow the change of the DC component. In this way, it is possible to improve the transient response to the change of the physical quantity signal PI and the followability to the change of the DC component without adding a new inertial sensor.

Specifically, as described later with reference to FIG. 3, the noise estimation unit 110 estimates the observation noise σ _meas based on the input signal PI, and performs a gain process on the estimated observation noise σ _meas . Estimate the noise σ _sys .

In this way, the observation noise σ _meas and the system noise σ _sys can be dynamically changed according to the input signal PI, and the characteristics of the Kalman filter 120 can be controlled according to the input signal PI. In addition, since the system noise σ _sys can be adjusted by changing the gain of the gain processing, it is possible to control the Kalman filter 120 so as to have a desired characteristic.

More specifically, the noise estimation unit 110 performs gain processing on the observation noise σ _meas with a gain set based on the target cutoff frequency of the low-pass filter operation in the convergence state of the Kalman filter 120.

When time passes sufficiently after the Kalman filter 120 starts to operate, as shown in the following equation (14) etc., an internal variable such as the Kalman gain converges to a constant value, and the Kalman filter 120 is a filter including low pass filter characteristics Converge on the characteristic. In this convergence state, the cutoff frequency f _c of the low pass filter is determined by the gain GA1 as shown in the following equations (17) to (19). Conversely, by setting the gain GA1 to obtain the desired cutoff frequency f _c , when the Kalman filter 120 is in the convergence state, a low pass filter of the desired cutoff frequency f _c Characteristics can be obtained.

  In the present embodiment, by using the Kalman filter 120, it is possible to obtain a very low cutoff frequency of, for example, about 0.1 mHz in the convergence state. As a result, only very low frequency components can be extracted as DC components DCQ from the physical quantity signal PI of the gyro sensor, and by subtracting the DC components DCQ, motion signal components can be output without losing the low frequency band. It can be output as the signal PQ.

As will be described later with reference to FIG. 3, the noise estimation unit 110 estimates the observation noise σ _meas by subjecting the input signal PI to the mean square and performing a limiter process on the signal after the mean square.

For example, in FIG. 3, since smoothing is performed with a very low frequency low pass filter 133 at the time of root mean square to remove the signal component of motion from the physical quantity signal PI, the root mean square noise component can be left. As described above, the observation noise σ _meas can be extracted from the input signal PI by squaring the input signal PI.

If the observation noise σ _meas becomes zero, the Kalman gain g (k) = 1, as can be seen from the following equation (8), and the estimated value x (k), as can be understood from the following equation (9) As the observation value y (k), the Kalman filter 120 outputs the observation value y (k) as it is. However, in practice, the observation noise σ _meas is never zero. In this respect, in the present embodiment, the observation noise σ _meas can be limited to a value larger than zero by limiter processing. For example, since the quantization noise when A / D-converting the output signal TQ of the gyro sensor is considered as the minimum observation noise, a value obtained by estimating the quantization noise may be set as the limit value.

Further, as described later with reference to FIG. 4, the noise estimation unit 110 performs high-pass filter processing on the input signal PI, and performs processing of increasing the observation noise σ _meas based on the signal after the high-pass filter processing. That is, the second estimation unit 140, by adding the signals generated from the input signal PI after high-pass filtering the observation noise sigma _meas, increase the observed noise sigma _meas.

When the gyro sensor detects a motion, the value of the input signal PI fluctuates and becomes not DC, so the reliability of the observed value decreases for the purpose of extracting the DC component. At this time, by increasing the observation noise σ _meas based on the input signal PI after the high-pass filter processing, the Kalman gain g (k) can be reduced as can be understood from the following equation (8). As understood from the following equation (9), when the Kalman gain g (k) becomes smaller, the Kalman filter 120 trusts the past estimated value x ⁻ (k). That is, when the physical quantity signal PI of the gyro sensor is not DC, the reliability of the physical quantity signal PI which is the observed value is lowered, and the past estimated value considered to be close to the true value can be output as the DC component DCQ.

Specifically, the noise estimation unit 110 performs peak hold processing on the signal after the high-pass filter processing, and increases the observation noise σ _meas based on the signal after the peak hold processing.

  As described above, when the gyro sensor detects a motion, the estimation accuracy of the DC component can be improved by making the Kalman filter 120 rely on the past estimated value. At this time, the DC component can be more accurately estimated by performing the peak hold process.

For example, in FIG. 4, the signal after high-pass filter processing is subjected to the square average, and the peak hold processing is performed on the signal after the square average. As shown in FIG. 5, when the signal after the high-pass filter processing is zero, the signal after the root mean square also becomes zero. Increasing the observation noise sigma _meas Based on such signal, the signal becomes the observation noise sigma _meas is at zero does not increase, as can be seen from the following equation (8), Kalman gain g (k) is growing. Then, as understood from the following equation (9), the Kalman filter 120 may rely on the unreliable observed value y (k), and an incorrect DC component may be output.

In this respect, by performing the peak hold process in the present embodiment, it is possible to suppress that the signal after the mean square becomes zero. Then, by increasing the observation noise σ _meas using the signal, it is possible to improve the estimation accuracy of the DC component by making the Kalman filter 120 reliable in the past estimated value when the movement of the gyro sensor is detected.

2. First Detailed Configuration of Signal Processing Device Next, the detailed configuration and operation of the signal processing device will be described. FIG. 3 shows a first detailed configuration example of the signal processing device 100. As shown in FIG. The signal processing apparatus 100 includes a noise estimation unit 110, a Kalman filter 120, a low pass filter 171, a delay unit 172, a low pass filter 173, and a subtraction unit 121.

The Kalman filter 120 extracts a DC component by a linear Kalman filter, using the input signal input from the delay unit 172 as an observation value. The basic formulas of the linear Kalman filter are the following formulas (1) to (5).

The above equations (1) and (2) are equations of time update (prediction process), and the above equations (3) to (5) are equations of observation update (observation process). k represents a discrete time, and one time update and one observation update are performed each time k goes forward. x (k) is an estimate of the Kalman filter 120, x ^- (k) is the pre-estimated value predicted prior to obtaining observations. P (k) is the error covariance of the Kalman filter 120, P ^- (k) is the error covariance predicted prior to obtaining observations. y (k) is an observation value. σ _sys (k) is system noise, and σ _meas (k) is observation noise. A is a coefficient matrix, and b and c are coefficient vectors. A, b and c depend on the state space model that models the system.

Since this embodiment is a linear linear model in the present embodiment, A = b = c = 1, and the above equations (1) to (5) become the following equations (6) to (10).

The Kalman filter 120 stores the estimated value x (k-1) and the error covariance P (k-1) updated at the previous time k-1. Then, at the current time k, the observation value y (k), the observation noise σ _meas (k) and the system noise σ _sys (k) are received, and using them, the time update and observation of the above equations (6) to (10) Update is performed, and the estimated value x (k) is output as a DC component.

  The low pass filter 171 is a decimation filter that lowers the frequency (for example, 15 kHz) of the physical quantity signal from the gyro sensor to the operating frequency (for example, 100 Hz) of the Kalman filter 120. The operating frequency is the input sampling frequency of the Kalman filter 120, which is different from the frequency of the internal operation. This filter can prevent the aliasing of the signal input to the Kalman filter 120. Since a steep attenuation characteristic is not required, for example, a first-order low pass filter is used. The cutoff frequency is set to, for example, about 1/4 to 1/6 of the operating frequency of the Kalman filter 120.

The delay unit 172 delays the input signal of the Kalman filter 120 by the same time as the time delayed by the calculation of the noise estimation unit 110. For example, when the noise estimation unit 110 delays by one step, the delay unit 172 also delays the input signal by one step, so that the timing of the observation noise σ _meas and the system noise σ _sys can be matched with the timing of the input signal.

  The low pass filter 173 performs a filtering process on the final output of the Kalman filter 120. Basically, the Kalman filter 120 should output a DC value, so the low pass filter 173 is set to a low cutoff frequency (eg, 0.1 Hz). The low pass filter 173 is an interpolation filter that upsamples the operating frequency (for example, 100 Hz) of the Kalman filter 120 to the frequency (for example, 15 kHz) of the physical quantity signal of the gyro sensor.

The noise estimation unit 110 generates a dynamically changing observation noise σ _meas and a system noise σ _sys based on the input signal PI processed by the low pass filter 171. Specifically, the noise estimation unit 110 includes a first estimation unit 130. The first estimation unit 130 includes a high pass filter 131, a square processing unit 132, a low pass filter 133, a limiter 134, and a gain processing unit 135.

The high pass filter 131 receives the signal from the low pass filter 171 and removes the DC component from the signal. Since the root mean square is performed in the latter stage, removing the DC component can prevent the DC component from being squared and becoming an error of the observation noise σ _meas .

  The square processing unit 132 squares the signal from the high pass filter 131. The low pass filter 133 filters the signal squared by the square processing unit 132 to obtain a square average. The noise component of the signal is extracted by this root mean square. Basically, the output is a DC value and the cutoff frequency is set to a very low value. For example, the cutoff frequency is set sufficiently lower than the frequency of the motion signal component detected by the gyro sensor. By doing this, only noise components included in the signal can be extracted.

The limiter 134 performs limit processing on the signal from the low pass filter 133, and outputs the signal after the limit processing as a variance (square of the observation noise σ _meas ). Specifically, when the signal from low pass filter 133 is below the lower limit, the output is limited to the lower limit, and when the signal from low pass filter 133 is larger than the lower limit, the signal is output as it is. . An A / D conversion unit (A / D conversion circuit 66 in FIG. 12) is provided at the front stage of the signal processing apparatus 100. Since the noise component can never fall below the quantization noise of the A / D converter, a lower limit value of, for example, 0.25 degit is provided to the output of the low pass filter 133.

The gain processing unit 135 multiplies the variance of the observation noise σ _meas from the limiter 134 by a constant gain GA 1 to obtain the variance (square) of the system noise σ _sys . The gain GA1 is set as shown in the following equation (19). The derivation method of the following equation (19) will be described below.

First, the relationship between the observation noise σ _meas and the system noise σ _sys in a state where a sufficient time has elapsed is obtained. The state in which the time has sufficiently passed may be assumed to be k = k, and assuming that the prior error covariance P ⁻ (k) converges to a constant value, the following equation (11) holds. Pre-error covariance P ^- the convergence value of (k) is set to p _0.

When the above equation (11) is applied to the above equations (7) and (10), the following equation (12) is obtained.

Further, when the above equation (11) is applied to the above equation (8), the following equation (13) is obtained.

When the above equations (12) and (13) are solved for the Kalman gain g (k) as simultaneous equations, the following equation (14) is obtained. In the following equation (14), the Karman gain g (k) in the convergence state k = ∞ is g. In the approximation on the right side, it is assumed that σ _sys << σ _meas holds because the passband is very low in the convergence state of the Kalman filter 120.

As understood from the above equation (14), the Kalman gain g becomes a ratio of the system noise σ _sys and the observation noise σ _meas in the convergence state. That is, it can be seen that σ _sys ² = g ² σ _meas ² and gain GA 1 = g ² . If the relationship between the desired filter characteristic for extracting the DC component and the Kalman gain g is known, the gain GA1 can be set so as to obtain the desired filter characteristic.

In order to know the relationship between the desired filter characteristics and the Kalman gain g, the transfer characteristics of the Kalman filter 120 in the convergence state are determined. From the above equations (6) and (9), the transfer function H (z) becomes the following equation (15). The Kalman filter 120 is characterized by continuing to change the configuration to the theoretically optimal filter based on the current estimated error, and does not always follow the transfer function of the following equation (15). The following equation (15) is the final transfer function when time passes.

When the bilinear transformation is applied to the above equation (15), the following equation (16) is obtained.

Looking at the denominator of equation (16) above, it can be seen that the transfer function includes low-pass filter characteristics. The cut-off frequency f _c of the low pass filter characteristic is the following equation (17). f _s is the sampling frequency (operating frequency) of the Kalman filter 120.

When the above equation (17) is solved for the Kalman gain g, the following equation (18) is obtained. In the approximation of the right side of the following equation (18), f _c << f _s .

From the above equation (18), the gain GA1 = ^{g 2} are determined in the following equation (19).

In the above equation (19), a desired cut-off frequency (target cut-off frequency) to be finally obtained in the convergence state is set to _fc . By setting such a gain GA1, the Kalman filter 120 acts as a low pass filter as shown in the above equation (16) when a sufficient time has elapsed, and the cutoff frequency tries to converge on f _c .

3. Second Detailed Configuration of Signal Processing Device FIG. 4 shows a second detailed configuration example of the signal processing device 100. As shown in FIG. The signal processing apparatus 100 includes a noise estimation unit 110, a Kalman filter 120, a low pass filter 171, a delay unit 172, a low pass filter 173, and a subtraction unit 121. The same components as those in the first detailed configuration example are denoted by the same reference numerals, and the description will be appropriately omitted.

The noise estimation unit 110 dynamically changes the observation noise σ _meas based on the first estimation unit 130 described above and the AC component (motion component) extracted from the physical quantity signal of the gyro sensor. including.

  Specifically, the second estimation unit 140 includes a peak hold unit 141 and an addition unit 142.

  The peak hold unit 141 receives the AC component signal that has passed through the high pass filter 131 and the square processing unit 132 of the first estimation unit 130, and peak holds the signal. The high-pass filter 131 needs to have a cut-off frequency not too low because the response is important in order to pass the motion signal detected by the gyro sensor. For example, the cutoff frequency is set to about 10 Hz. The peak hold process is, for example, a process of comparing a past signal and a current signal and holding the larger signal. As shown in FIG. 5, when the current signal does not exceed the signal being held for a certain period (hold period), the hold is stopped and the current signal is output.

The addition unit 142 adds the signal from the peak hold unit 141 and the observation noise σ _meas from the first estimation unit 130, and outputs the signal after the addition to the Kalman filter 120. As the movement detected by the gyro sensor is larger, the signal from the peak hold unit 141 is also larger. _Therefore , as the movement is larger, the observation noise σ _meas increases. As the observation noise σ _meas increases, the Kalman gain g (k) decreases as can be seen from the above equation (8), and the weight of the observed value y (k) is lowered to estimate the estimated value as can be seen from the above equation (9) x (k) can be calculated. Thereby, the influence of the observation value y (k) is reduced as the AC component of the motion is larger, and the DC component with higher accuracy can be extracted.

The reason for performing peak hold will be described. As shown in FIG. 5, when the gyro sensor detects motion and the output of the high-pass filter 131 vibrates violently, the output of the square processing unit 132 obtained by squaring the output has a moment when it drops to zero. I know the thing. At that moment, the weight of the observation value y (k) can not be reduced because it is not added to the observation noise σ _meas despite the AC component in the signal. Therefore, the observed value y (k) on which the AC component is applied affects the estimated value x (k) to some extent, and the estimation accuracy of the DC component is lowered.

In this regard, peak hold processing can prevent the signal to be added to the observation noise σ _{meas from} becoming zero. The hold time may be set, for example, by the number of operation cycles of the Kalman filter 120, and an appropriate time may be determined based on experimental results and the like. For example, several tens of cycles are set.

4. Third Detailed Configuration of Signal Processing Device FIG. 6 shows a third detailed configuration example of the signal processing device 100. As shown in FIG. The signal processing apparatus 100 includes a noise estimation unit 110, a Kalman filter 120, a low pass filter 171, a delay unit 172, a low pass filter 173, a subtraction unit 175, an addition unit 176, a storage unit 177, a reading unit 178, a comparator 180, and a subtraction unit 121. including. The same components as those in the first and second detailed configuration examples will be assigned the same reference numerals and descriptions thereof will be omitted as appropriate.

  The storage unit 177 stores an offset table in which the temperature and the DC designated value (approximate DC value) as illustrated in FIG. 7 are associated with each other. The designated DC value is a central value of deviation obtained by approximating the temperature characteristic by a quadratic function of temperature. The storage unit 177 includes, for example, a non-volatile memory (for example, an EPROM). The reading unit 178 refers to the offset table based on the detection signal TS from the temperature sensor 190, and outputs a DC designated value corresponding to the detected temperature. Subtraction unit 175 subtracts the DC designation value from the output signal of low pass filter 171, and outputs the signal after subtraction to delay unit 172 and noise estimation unit 110. The addition unit 176 adds the output signal of the Kalman filter 120 and the DC designation value, and outputs the signal after the addition to the low pass filter 173.

  As described above, the Kalman filter 120 and the noise estimation unit 110 perform processing on the input signal after subtracting the DC designation value. This corresponds to pre-approximating the expected DC value. By doing this, it is possible to improve the convergence of DC estimation, to prevent estimation with far apart DC values even in a situation where DC estimation with severe fluctuations is difficult, and to improve reliability. Once estimated with far-off DC values, the time constant of the Kalman filter 120 (eg, the inverse of the cutoff frequency of 0.1 mHz) may be very long, so it may take an excessive amount of time to return to the original estimate. Although this is possible, this can be prevented in the present embodiment.

The noise estimation unit 110 includes the first estimation unit 130 and the second estimation unit 140 described above, and the third estimation unit 150 for dynamically changing the observation noise σ _meas according to the degree of separation from the DC designated value. Including.

  Specifically, the third estimation unit 150 includes an absolute value calculation unit 151, a gain processing unit 152, a limiter 153, and a multiplication unit 154.

  The storage unit 177 further stores the designated value d used by the gain processing unit 152 and the maximum deviation value (dotted line in FIG. 7) of the designated DC value used by the comparator 180. The maximum deviation value is, for example, a value having no temperature characteristic. For example, a plurality of samples of the gyro sensor are measured, and the DC designation and the maximum deviation value are set based on the results. In order to use the maximum deviation value at the subsequent stage of the absolute value calculation unit 151, in actuality, the absolute value of the maximum deviation value may be stored. Although any value can be set as the designated value d, it is identical to, for example, the maximum deviation value. Hereinafter, the case where the designated value d is the maximum deviation value will be described as an example.

The absolute value calculating unit 151 calculates the absolute value of the signal from the subtracting unit 175. The gain processing unit 152 multiplies the absolute value by the gain GA2 of the following expression (20). The limiter 153 outputs the signal as it is when the signal after multiplication is larger than 1 and limits the output to 1 when the signal after multiplication is 1 or less. The multiplication unit 154 multiplies the signal from the limiter 153 by the observation noise σ _meas from the second estimation unit 140.

That is, since the absolute value becomes larger than d when the physical quantity signal of the gyro sensor is out of the deviation range, it becomes larger than 1 when it is multiplied by the gain GA2. In this case, the limiter 153 outputs a value larger than 1 and the observation noise σ _meas increases. On the other hand, when the physical quantity signal is within the deviation range, the absolute value is smaller than d, so when it is multiplied by the gain GA2, it becomes 1 or less. In this case, the limiter 153 outputs 1 and the observation noise σ _meas does not increase.

As the observation noise σ _meas increases, the estimated value x ⁻ (k) is trusted from the above equations (8) and (9), so the weight of the observed value y (k) decreases outside the deviation range. Conversely, in the deviation range, the weight of the observation value y (k) increases, and in the deviation range, the DC component of the observation value y (k) is intensively searched. In this manner, DC estimation is performed mainly by searching within the deviation range in which the true value of DC seems to exist, and lowering the reliability of the observed value y (k) that is considered not to be the true value outside the deviation range. Can improve the accuracy of

  As described above, since the present embodiment performs DC estimation intensively within the range of deviation, if the true value of DC is out of the range, the DC estimation speed becomes extremely slow, and disturbance (signal (Even for large inputs), the DC estimate fluctuates quickly. Therefore, it is necessary not to make the maximum deviation value too small so that the true value falls within the deviation, and to designate it in consideration of aging.

  Next, the comparator 180 (monitoring unit) will be described. The comparator 180 monitors the movement component of the physical quantity signal of the gyro sensor, and outputs a flag FL1 (stop signal) indicating the presence or absence of the movement. Specifically, the signal from the absolute value calculation unit 151 is compared with the threshold with the maximum deviation value as the threshold, and the flag FL1 is activated when the signal is larger than the threshold. That is, when a movement larger than the deviation range occurs, the flag FL1 becomes active. The flag FL1 is input to the Kalman filter 120 and the low pass filter 133 of the first estimation unit 130.

  When motion is added to the gyro sensor and the input signal changes significantly, the accuracy of DC estimation decreases, and the error covariance P (k) corresponding to the estimation error increases. Then, the Kalman gain g (k) approaches 1 as can be seen from the above equation (8), and the observed value y (k) affected by the motion is trusted as can be seen from the above equation (9).

Therefore, when the flag FL1 is active, the Kalman filter 120 performs only the estimation of the estimated value x (k) and does not update the error covariance P (k). Specifically, the processes of the equations (6) to (9) are executed, and the process of the equation (10) is not executed. As a result, it is possible to increase the reliability of the past estimated value x ⁻ (k) that is less affected by motion, and to improve the estimation accuracy of the DC component.

The low pass filter 133 stops the low pass filter operation when the flag FL1 is active, and continues to output an output value when the flag FL1 becomes active from inactive. As a result, only the signal within the deviation range is low-pass filtered, and it is possible to prevent extraction of the imprecise observation noise σ _meas by the signal affected by the motion.

5. Fourth Detailed Configuration of Signal Processing Device FIG. 8 shows a fourth detailed configuration example of the signal processing device 100. As shown in FIG. The signal processing apparatus 100 includes a noise estimation unit 110, a Kalman filter 120, a low pass filter 171, a delay unit 172, a low pass filter 173, a subtraction unit 175, an addition unit 176, a storage unit 177, a reading unit 178, a comparator 180, and a subtraction unit 121. including. The same components as those in the first to third detailed configuration examples will be assigned the same reference numerals and descriptions thereof will be omitted as appropriate.

  The noise estimation unit 110 restores the Kalman filter 120 from the convergence state to the estimation state when there is a temperature change and the first estimation unit 130, the second estimation unit 140, and the third estimation unit 150 described above. Estimation unit 160 of FIG.

  As shown in FIG. 9, the DC offset of the gyro sensor has a temperature characteristic, and its approximate value is stored as an offset table, subtracted by the subtraction unit 175, and input to the Kalman filter 120. The approximation does not exactly match the temperature characteristics, and some individuals have dips in the temperature characteristics, so the temperature characteristics also remain as a result of subtraction. As shown in FIG. 10, when the temperature changes with the passage of time, the input of the Kalman filter 120 also changes according to the remaining temperature characteristic.

When the DC estimation of the Kalman filter 120 converges, it has almost the same characteristics as the low pass filter of the target cutoff frequency f _c as described in the first detailed configuration example. Under this condition, when the DC value of the input signal changes due to a temperature change, the DC value before the temperature change continues to be output for a while because the cut-off frequency f _c is low. The time is about the inverse of the specified target cutoff frequency f _c . That is, the time is that the Kalman filter 120 can not follow the change of the DC value due to the temperature change.

Therefore, in the present embodiment, the Kalman filter 120 is returned to the estimated state by dynamically changing the system noise σ _sys according to the temperature change, and the followability to the change of the DC value is improved. Specifically, the fourth estimation unit 160 includes a delay unit 161, a subtraction unit 162, a low pass filter 163, a gain processing unit 164, a square processing unit 165, a multiplication unit 166, and an addition unit 167.

The delay unit 161 and the subtraction unit 162 obtain the difference between the detection signal TS at time k of the temperature sensor 190 and the detection signal TS at time k−1 immediately before. The low pass filter 163 smoothes the difference. When the noise of the detection signal TS of the temperature sensor 190 is large, or when the sampling frequency of the detection signal TS is lower than the operating frequency of the Kalman filter 120, the fourth estimation unit 160 may make a discrete change. The system noise σ _sys output by changes suddenly and significantly. This change greatly affects the degree of convergence of the Kalman filter 120. Therefore, an extremely low cutoff frequency (for example, 10 mHz) is set to the low pass filter 163. By this, it is possible to reliably remove the component to be the noise and to make the change of the system noise σ _sys gentle.

Using the sensitivity TSEN [digi / ° C.] of the temperature sensor 190 and the temperature change ΔT [° C./sec], the output LPF2 _OUT of the low-pass filter 163 is expressed by the following equation (21).

The gain processing unit 164 multiplies the signal from the low pass filter 163 by the gain GA3. The squaring processor 165 squares the signal after multiplication. The multiplication unit 166 multiplies the signal after the square and the observation noise σ _meas from the first estimation unit 130. Then, the addition unit 167 adds the signal after the multiplication and the system noise σ _sys from the first estimation unit 130, and outputs the signal after the addition to the Kalman filter 120.

The setting method of gain GA3 is demonstrated. First, the change of the input signal of the Kalman filter 120 when the temperature changes is approximated by a sine wave. As shown in FIG. 10, considering the slope by first order approximation, the following equation (22) is obtained. k [digi / ° C.] is a temperature variation coefficient of the physical quantity signal of the gyro sensor. Since ΔT [° C./sec] is a temperature change per second, kΔT is the slope of the change of the input signal. The slope is approximated by the slope 2πf of the sine wave of frequency f. Substituting the above equation (21) into ΔT, and solving for f, becomes the right side of the following equation (22).

Since the frequency of the change of the DC value is found to be f, the gain GA3 is determined so that the frequency passes through the Kalman filter 120. That is, when the frequency f in the above equation (22) is substituted as the cutoff frequency f _c into the above equation (18) in the converged state, the following equation (23) is obtained.

When the sensitivity of the gyro sensor is SEN [digi / dps] and the temperature coefficient is TCOEFF [dps / ° C.], the above equation (23) becomes the following equation (24).

By setting the gain GA3 as described above, it is possible to change the cutoff frequency f _c of the low-pass filter in the convergence state to the frequency f and return the Kalman filter 120 to the estimation state. At this time, since the cutoff is the frequency f, the change of the DC component of the input signal passes through at the last moment.

As described above, information as to how much the DC value changes due to temperature change can be given to the Kalman filter 120 as the system noise σ _sys , whereby the Kalman filter 120 can be used for the change of the DC component due to the temperature change. Can be made to follow.

6. Electronic Device, Gyro Sensor FIG. 11 shows a configuration example of a gyro sensor 510 (a sensor in a broad sense) including the signal processing device 100 of the present embodiment and an electronic device 500 including the gyro sensor 510. The electronic device 500 and the gyro sensor 510 are not limited to the configuration shown in FIG. 11, and various modifications may be made such as omitting some of the components or adding other components. Further, as the electronic device 500 of the present embodiment, various devices such as a digital camera, a video camera, a smartphone, a mobile phone, a car navigation system, a robot, a game machine, a watch, a health appliance, or a portable information terminal can be assumed.

  The electronic device 500 includes a gyro sensor 510 and a processing unit 520. In addition, a temperature sensor 190, a memory 530, an operation unit 540, and a display unit 550 can be included. The processing unit 520 (CPU, MPU, etc.) performs control of the gyro sensor 510 and the like and overall control of the electronic device 500. The processing unit 520 also performs processing based on angular velocity information (physical quantity in a broad sense) detected by the gyro sensor 510. For example, processing for camera shake correction, attitude control, GPS autonomous navigation, etc. is performed based on angular velocity information. The memory 530 (ROM, RAM, etc.) stores control programs and various data, and functions as a work area and a data storage area. The operation unit 540 is for the user to operate the electronic device 500, and the display unit 550 displays various information to the user.

  The gyro sensor 510 includes the vibrator 10 and the detection device 20. The vibrator 10 (physical quantity transducer in a broad sense) of FIG. 11 is a tuning fork type piezoelectric vibrator formed of a thin plate of a piezoelectric material such as quartz, and includes driving vibrators 11 and 12 and a detecting vibrator 16. , 17 have. Drive terminals 2 and 4 are provided to the drive vibrators 11 and 12, and detection terminals 6 and 8 are provided to the detection vibrators 16 and 17.

  The detection device 20 includes a drive circuit 30, a detection circuit 60, and a signal processing device 100. The drive circuit 30 outputs a drive signal (drive voltage) to drive the vibrator 10. Then, a feedback signal is received from the vibrator 10 to excite the vibrator 10. The detection circuit 60 receives a detection signal (detection current, charge) from the vibrator 10 driven by the drive signal, and a desired signal (physical quantity signal, Coriolis force signal according to the physical quantity applied to the vibrator 10 from the detection signal. ) Is detected (extracted).

  Specifically, an AC drive signal (drive voltage) from the drive circuit 30 is applied to the drive terminal 2 of the drive vibrator 11. Then, the drive vibrator 11 starts to vibrate by the reverse voltage effect, and the drive vibrator 12 also starts to vibrate by the tuning fork vibration. At this time, the current (electric charge) generated by the piezoelectric effect of the drive vibrator 12 is fed back from the drive terminal 4 to the drive circuit 30 as a feedback signal. Thereby, an oscillation loop including the vibrator 10 is formed.

  When the drive vibrators 11 and 12 vibrate, the detection vibrators 16 and 17 vibrate at the vibration speed v in the direction shown in FIG. Then, the current (charge) generated by the piezoelectric effect of the detection vibrators 16 and 17 is output from the detection terminals 6 and 8 as detection signals (first and second detection signals). Then, the detection circuit 60 receives the detection signal from the vibrator 10 and detects a desired signal (desired wave) which is a signal corresponding to the Coriolis force. That is, when the vibrator 10 (gyro sensor) rotates around the detection axis 19, the Coriolis force Fc is generated in the direction orthogonal to the vibration direction of the vibration velocity v. For example, assuming that the angular velocity when rotating about the detection axis 19 is ω, the mass of the oscillator is m, and the oscillation velocity of the oscillator is v, the Coriolis force is expressed as Fc = 2 m · v · ω. Accordingly, the rotational angular velocity ω of the gyro sensor can be obtained by detecting the desired signal which is a signal corresponding to the Coriolis force. Then, using the obtained angular velocity ω, the processing unit 520 can perform various processes for camera shake correction, attitude control, GPS autonomous navigation, and the like.

  In addition, although the example in case the vibrator | oscillator 10 is a tuning fork type is shown in FIG. 11, the vibrator | oscillator 10 of this embodiment is not limited to such a structure. For example, it may be T-shaped or double T-shaped. The piezoelectric material of the vibrator 10 may be other than quartz.

7. Detection Device FIG. 12 shows a configuration example of the detection device 20 of the present embodiment. The detection device 20 includes a drive circuit 30, a detection circuit 60, and a signal processing device 100 (digital processing unit).

  Drive circuit 30 includes an amplifier circuit 32 to which feedback signal DI from vibrator 10 is input, a gain control circuit 34 for performing automatic gain control, and a drive signal output circuit 36 for outputting drive signal DQ to vibrator 10. . It also includes a synchronization signal output circuit 38 that outputs synchronization signal SYC to detection circuit 60. The configuration of drive circuit 30 is not limited to that shown in FIG. 12, and various modifications may be made such as omitting some of these components or adding other components.

  The amplification circuit 32 (I / V conversion circuit) amplifies the feedback signal DI from the vibrator 10. For example, the signal DI of the current from the vibrator 10 is converted into a signal DV of voltage and output. The amplification circuit 32 can be realized by a capacitor, a resistance element, an operational amplifier or the like.

  The drive signal output circuit 36 outputs a drive signal DQ based on the signal DV amplified by the amplifier circuit 32. For example, the drive signal output circuit 36 outputs a rectangular wave (or sine wave) drive signal. The drive signal output circuit 36 can be realized by a comparator or the like.

  The gain control circuit 34 (AGC) outputs the control voltage DS to the drive signal output circuit 36 to control the amplitude of the drive signal DQ. Specifically, the gain control circuit 34 monitors the signal DV to control the gain of the oscillation loop. For example, in the drive circuit 30, in order to keep the sensitivity of the gyro sensor constant, it is necessary to keep the amplitude of the drive voltage supplied to the vibrator 10 (drive vibrator) constant. For this reason, a gain control circuit 34 for automatically adjusting the gain is provided in the oscillation loop of the drive vibration system. The gain control circuit 34 automatically adjusts the gain variably so that the amplitude of the feedback signal DI from the vibrator 10 (vibration speed v of the vibrator) becomes constant.

  The synchronization signal output circuit 38 receives the signal DV amplified by the amplification circuit 32, and outputs a synchronization signal SYC (reference signal) to the detection circuit 60. The synchronization signal output circuit 38 performs a binarization process on the sine wave (AC) signal DV to generate a rectangular wave synchronization signal SYC, and a phase adjustment circuit (phase shift circuit for performing phase adjustment of the synchronization signal SYC). It can be realized by a phaser or the like.

  The detection circuit 60 includes an amplification circuit 62, a synchronous detection circuit 64, and an A / D conversion circuit 66. The amplification circuit 62 receives the first and second detection signals IQ1 and IQ2 from the vibrator 10, and performs signal amplification and charge-voltage conversion. The synchronous detection circuit 64 performs synchronous detection based on the synchronous signal SYC from the drive circuit 30. The A / D conversion circuit 66 performs A / D conversion of the signal after synchronous detection.

  As described above, the signal processing apparatus 100 extracts the DC offset (DC component) from the desired signal, and removes the DC offset from the desired signal. Also, the signal processing apparatus 100 may perform various digital signal processing. For example, the signal processing apparatus 100 performs band-limited digital filter processing according to the application of a desired signal, and digital filter processing for removing noise generated by the A / D conversion circuit 66 and the like. In addition, digital correction processing such as gain correction (sensitivity adjustment) and offset correction is performed. The signal processing apparatus 100 can be realized by, for example, a DSP (Digital Signal Processor).

8. Mobile Body FIG. 13 shows an example of a mobile body including the detection device 20 of the present embodiment. The detection device 20 of the present embodiment can be incorporated into various mobile objects such as, for example, a car, an airplane, a motorcycle, a bicycle, or a ship. The movable body is, for example, a device / device that moves on the ground, in the sky or in the sea, provided with a drive mechanism such as an engine or a motor, a steering mechanism such as a steering wheel or a rudder, and various electronic devices.

  FIG. 13 schematically shows a car 206 as an example of a mobile. The automobile 206 incorporates a gyro sensor 510 (sensor) having the vibrator 10 and the detection device 20. The gyro sensor 510 detects the posture of the vehicle body 207. The physical quantity signal of the gyro sensor 510 is supplied to the vehicle body posture control device 208. The vehicle attitude control device 208 controls the hardness of the suspension or controls the brakes of the individual wheels 209 in accordance with the attitude of the vehicle body 207, for example. In addition, such attitude control can be used in various mobile bodies such as biped robots, aircraft, and helicopters. A gyro sensor 510 can be incorporated to realize the attitude control.

  It should be understood by those skilled in the art that although the present embodiment has been described in detail as described above, many modifications can be made without departing substantially from the novel matters and effects of the present invention. Accordingly, all such modifications are intended to be included within the scope of the present invention. For example, in the specification or the drawings, the terms described together with the broader or synonymous different terms at least once can be replaced with the different terms anywhere in the specification or the drawings. Further, all combinations of the present embodiment and the modifications are also included in the scope of the present invention. Further, the signal processing device, the detection device, the sensor, the electronic device, the configuration / operation of the moving body, etc. are not limited to those described in the present embodiment, and various modifications can be made.

2, 4 drive terminals, 6, 8 detection terminals, 10 vibrators, 11, 12 drive vibrators,
15 physical quantity transducers, 16 and 17 transducers for detection, 19 detection axes,
20 detection devices, 30 drive circuits, 32 amplification circuits, 34 gain control circuits,
36 drive signal output circuit, 38 synchronization signal output circuit, 60 detection circuit,
62 amplifier circuit, 64 synchronous detection circuit, 66 A / D converter circuit,
100 signal processing units, 110 noise estimation units, 120 Kalman filters,
121 subtraction unit, 130 first estimation unit, 131 high pass filter,
132 low pass filter, 133 low pass filter, 134 limiter,
135 gain processing unit, 136 addition unit, 140 second estimation unit,
141 peak hold unit, 142 addition unit, 150 third estimation unit,
151 absolute value calculation unit, 152 gain processing unit, 153 limiter,
154 multiplication unit, 160th fourth estimation unit, 161 delay unit, 162 subtraction unit,
163 low pass filter, 164 gain processor, 166 multiplier,
167 adder, 171 low pass filter, 172 delay,
173 low pass filter, 175 subtraction unit, 176 addition unit, 177 storage unit,
178 readout units, 180 comparators (monitoring units), 190 temperature sensors,
206 car, 207 vehicle body, 208 vehicle body attitude control device, 209 wheel,
500 electronics, 510 gyro sensors, 520 processors,
530 memory, 540 operation unit, 550 display unit,
DCQ DC component, FL1 flag, GA1 to GA3 gain,
PI input signal (physical quantity signal), PQ output signal, σ _meas observation noise,
σ _sys system noise

Claims (11)

A noise estimation unit for generating a dynamically varying observation noise variance and system noise variance,
An input signal, the observation noise variance, and the system noise variance, and performing a first-order Kalman filtering process on the input signal to extract a DC component of the input signal as an estimated value of the Kalman filtering process ; ,
Only including,
The signal processing apparatus characterized in that the Kalman filter processing includes time update processing of an error covariance using the system noise dispersion, and observation update processing of a Kalman gain using the observation noise dispersion . In the signal processing device according to claim 1,
The noise estimation unit
A signal processing apparatus that generates the observation noise variance and the system noise variance based on the input signal. In the signal processing device according to claim 1 or 2 ,
The noise estimation unit
A signal processing apparatus characterized in that the system noise variance is generated by generating the observation noise variance based on the input signal and performing gain processing on the generated observation noise variance . In the signal processing device according to claim 3 ,
The noise estimation unit
A signal processing apparatus characterized in that the observation noise variance is generated by performing a mean square of the input signal and performing a limiter process on the signal after the mean square. In the signal processing device according to claim 3 or 4 ,
The noise estimation unit
A signal processing apparatus characterized in that the gain processing is performed on the observation noise dispersion by a gain set based on a target cutoff frequency of a low pass filter operation in a convergence state of the Kalman filter. The signal processing apparatus according to any one of claims 1 to 5 .
The noise estimation unit
A signal processing apparatus characterized by performing high pass filter processing on the input signal and performing processing to increase the observation noise variance based on the signal after the high pass filter processing. In the signal processing device according to claim 6 ,
The noise estimation unit
A signal processing apparatus characterized by performing peak hold processing on the signal after the high-pass filter processing and increasing the observation noise variance based on the signal after the peak hold processing. A drive circuit for driving a physical quantity transducer;
A detection circuit that receives a detection signal from the physical quantity transducer and detects a physical quantity signal according to the physical quantity;
The signal processing apparatus according to any one of claims 1 to 7 , wherein the DC component is extracted using the physical quantity signal as the input signal.
A detection device characterized in that it comprises: A detection device according to claim 8 ;
The physical quantity transducer;
A sensor characterized by including. An electronic apparatus comprising the signal processing device according to any one of claims 1 to 7 . A mobile unit comprising the signal processing device according to any one of claims 1 to 7 .