JPH03121438A - Shaking signal processor for camera - Google Patents

Abstract

PURPOSE:To eliminate high frequency components (noise) while less phase lag is caused by changing the cutting-off frequency of a low-pass filter according to the average frequency of a shaking signal. CONSTITUTION:A processor is provided with an accelation detecting means 1, the low-pass filter 11, means 19 and 20 changing the cutting-off frequency of the low-pass filter 11, and a means 3 monitoring a current cutting-off frequency. The cutting-off frequency of the low-pass filter 11 is determined based on the frequency of the shaking signal detected by the accelation detecting means 1. Here, the high frequency components such as the mechanical noise of a camera is eliminated from the accelation signal detected by the accelation detecting means 1, the average frequency of only signals related to shake is obtained, and the cutting-off frequency of the low-pass filter 11 varies according to the obtained average frequency so that the high frequency components can be eliminated while the less phase lag is caused by the low-pass filter 11. Thus, the phase lag of the shaking signal can be lessened and the high frequency components are eliminated.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a camera shake signal processing device, and more particularly to a camera shake signal processing device used for detecting the amount of camera shake.

[Prior Art] In order to detect the amount of camera shake, the acceleration signal detected by the acceleration detection means is integrated once to convert it into a velocity signal, and integrated twice to convert it into a displacement signal, thereby detecting the amount of camera shake. Means for calculating is already known. in this case,
Acceleration sensors are mainly used as means for detecting acceleration, but they are difficult to detect because they are used to detect camera shake signals.

Sensors that can detect acceleration from C level are often used,
In particular, with regard to signal processing near C, methods using bypass filters are being considered. By the way, based on the results of previous studies and experience, it is said that the frequency of camera shake signals is approximately 30 Hz from C level, but the purpose of detecting camera shake signals is to detect higher frequency signals. Processing along these lines has not been carried out much. By integrating the acceleration signal, it is possible to cut out high frequency signals to some extent, but there is also a trade-off with the integration constant, so a great effect cannot be expected.

[Problems to be Solved by the Invention] Incidentally, the acceleration detection means used to detect camera shake signals must be able to detect signals from the D and C levels, so a sensor with fairly good accuracy and performance is required. used. However, the high accuracy of the acceleration sensor means that signals in a frequency range other than camera shake signals,
In other words, signals in the frequency range from about 30 Hz or higher to the resonance frequency unique to the acceleration sensor are detected without exception. In this case, the voltage corresponding to the acceleration output from the acceleration detection means is proportional to the acceleration, and the acceleration is proportional to the square of the frequency, so the higher the frequency, the higher the output voltage of the sensor becomes. There is a great possibility that the output will saturate from an acceptable level. Then, for signal processing, this acceleration signal is
Since the camera shake correction signal is formed by integrating once and converting it into a velocity signal and integrating twice and converting it into a displacement signal, if the acceleration signal is saturated before the integration, Since a saturated signal is integrated, it becomes impossible to accurately detect the amount of camera shake.

On the other hand, the acceleration detection means for detecting the camera shake signal is naturally built into the camera, and the camera shake signal is detected mainly before and after exposure. By the way, before and after this exposure, at least the shutter and mirror operate inside the camera, so considerable high-frequency mechanical noise is generated. Therefore, the acceleration detection means for detecting the camera shake signal also detects high-frequency mechanical noise that is unrelated to the camera shake signal, as described above.

Up until now, acceleration sensors have been mainly used as means for detecting camera shake signals, and the problem has been how to perform accurate detection in the frequency range of camera shake signals. If it is incorporated inside the mirror, at least the above-mentioned problems will occur. in this case,
The output of the acceleration detection means can be passed through a low-pass filter to actively remove high-frequency components such as mechanical noise, but if the cutoff frequency of the low-pass filter is increased, the phase delay will be reduced, but noise will be introduced. On the other hand, if you reduce the cutoff frequency of the low-pass filter, noise can be removed, but the phase delay will increase.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and is capable of removing high-frequency components (noise) with a small phase lag from the output of an acceleration detection means for detecting camera shake signals. Its purpose is to provide a signal processing device.

[Means and effects for solving the problem] The camera shake signal processing device of the present invention includes a shake detection means for detecting camera shake and outputting a shake signal, and at least a low-pass sensor for processing the shake signal. - signal processing means including a filter, and frequency identification means for identifying the blurring signal within the blurring frequency range from the output of the signal processing means;
average frequency calculation means for calculating the average frequency of the blur signal identified by the frequency identification means; and cutoff frequency changing means for changing the cutoff frequency of the low-pass filter based on the calculated average frequency. It is characterized by having.

〔Example〕

The present invention will be explained below with reference to illustrated embodiments.

First, before explaining the present invention in detail, the basic principle of the present invention will be explained below.

The camera shake signal processing device of the present invention includes an acceleration detection means for detecting a camera shake signal, a low-pass filter for processing the signal, and a cutoff frequency of the low-pass filter. and a means for monitoring the current cutoff frequency, and determining the cutoff frequency of the low-pass filter based on the frequency of the camera shake signal detected by the acceleration detection means. It is. Then, high-frequency components such as camera mechanical noise are removed from the acceleration signal detected by the acceleration detection means, the average frequency of only the signals related to blur is determined, and the phase delay is applied using a low-pass filter according to the average frequency. The cutoff frequency of the low-pass filter is made variable so that high-frequency components can be removed with as little as possible. Furthermore, when the blur frequency suddenly changes, the cutoff frequency of the low-pass φ filter is reset to a predetermined value. Next, one embodiment of the present invention will be described with reference to FIGS. 1 to 5.

FIG. 1 is a block diagram of a camera shake signal processing apparatus according to the present invention, which is characterized by changing the cutoff frequency of a low-pass filter in a signal processing system in accordance with a camera shake signal. In the figure, reference numeral 1 denotes an acceleration detection means for detecting a camera shake signal, 2 a signal processing means for performing amplification processing, filter processing, etc. on the signal output from the acceleration detection means 1, and 3 a signal processing means for performing the above-mentioned signal processing means. This is an integrating means that integrates the signal output from 2, converts it into a speed signal, and supplies it to an actuator and a control system (not shown).

Next, a means for identifying frequencies will be explained.

First, 4 is a differentiation circuit that differentiates the signal supplied from the signal processing means 2 and separates the frequencies in the acceleration signal, and 5 is only the signal component with a positive sign among the signals output from the differentiation circuit 4. The first comparison means to extract 6
is a second comparison means that extracts only negative signal components. 2 output from the 1st 2nd comparing means 5.6
The two signals are processed by the following fifth comparison means 7. and a sixth comparing means 8, where a comparison is made to remove camera mechanical noise, etc. that are not directly related to the camera shake signal.

The outputs of the fifth and sixth comparison means 7.8 are then applied to the first counting means 9.8. The signal sent to the second counting means 10 and output from the 5.6th comparing means 7.8 is counted based on the clock signal for counting timing to obtain frequency data. This 1.2 counting means 9.10
The signal output from the average frequency calculation means 1, which will be described later
It is now sent to 8th.

On the other hand, the signal output from the signal processing means 2 is sent to the differentiation circuit 4 at the same time as L.

It is also sent to P and F (low pass filter) 11.

This, P, Fil cuts the high frequency component in the normal acceleration signal.
The post-processing circuits, that is, the third comparison means 12 to the fourth counting means 17, are constructed in the same manner as the post-processing circuit of the differentiating circuit 4. That is, the 3.4th comparison means 12.13 is the first and second comparison means 5.6 and the seventh and eighth comparison means 1.
4.15 is the comparison means 7 and 8 of the above 5°6, and
Since the third and fourth counting means 16 and 17 have exactly the same functions as the first and second counting means 9 and 10, their explanations will be omitted here. In addition, Section 3.4
The signals outputted from the counting means 16 and 17, that is, the frequency data, are sent to the average frequency calculating means 18, which will be described later.

The signals inputted to the differentiating circuit 4 and the filters P and Fil are subjected to various signal conversions as described above, and then sent to the average frequency calculation means 18. The frequency data sent from the first to fourth counting means 9°, 10, 16, and 17 are sequentially input to the calculation means 18, and the average of the frequency data is calculated. Based on the average frequency calculated here,
The cutoff frequency change determination means 19 determines the cutoff frequency of the low-pass filter within the signal processing means 2. Then, in order to change to the cutoff frequency determined here, in the cutoff frequency changing means 20,
Signal processing means 2 by electrical means such as analog switches
The cutoff frequency is changed by changing the values of the elements constituting the low-pass filter.

FIG. 5 is a circuit diagram of the cutoff frequency changing means 20. In the figure, this changing means 20 is an analog switch AS AS AS3. AS4 and
This anal opening 1'2' switching switch ASl, AS2. AS3. OR gates OR, OR2, OR3, OR for determining whether the signal source for controlling AS4 is the cutoff frequency change determination means 19 or the cutoff frequency forced change determination means 22 described below.
4, 7nd gate AND AND AND3.
AND4 is constructed 1'2. And analog switch A S t
.

As2. AS3. When AS4 is controlled by the cutoff frequency change determining means 19, the output line p4 of the cutoff frequency forced change determining means 22 is set to "L" level, and the output line II5.16. fJ7. f18 are respectively set to the "H° level. Then, based on the results obtained by the average frequency calculation means 18, the cutoff frequency change determination means 1
9 calculates the optimum cutoff frequency and switches the resistors R1, R2, R3, and R4 to correspond to the frequency. Also, the analog switch ASL.

As2. AS8. When AS4 is controlled by the cut-off frequency forced change determination means 22 described later, the output line II4 of the means 22 is set to the "H# level," and analog switches are connected to the output lines N5, #6, N, and #8. It is designed to send out control signals.

Returning to FIG. 1, the timing and duration of average frequency calculation in the average frequency calculating means 18 and cutoff frequency changing means 20 are determined by the control unit 21. This control by the control unit 21 is performed in conjunction with the overall sequence of the camera, such as the shirt release of the camera. Further, the cutoff frequency forced change determination means 22 compares the frequency of the current camera shake signal and the cutoff frequency of the low-pass filter in the signal processing means 2, and determines that the state is inappropriate. if you do,
The cutoff frequency changing means 201 is a means for providing information. Information for this comparison/judgment is provided by the third counting means 16. Fourth counting means 17. and the cutoff frequency change determination means 19.

The functions and effects of this embodiment configured as described above will be explained below. In FIG. 1, an acceleration signal detected as a camera shake signal is amplified by a signal processing means 2 and a high frequency component is cut by a low-pass filter.

Then, the signal is normally sent to the integrating means 3 and converted into a speed signal or a displacement signal, which becomes a control signal for the camera shake correction actuator. The same signal inputted to the integrating means 3 is also inputted to the differentiating circuit 4, P, and Fil. Several specific examples of input waveforms are assumed as the waveforms of the signals input to the differentiating circuit 4, P, and Fil, and the response of this embodiment to each of these input waveforms will be explained in order below.

As an example of the first input waveform, as shown in FIG.
For example, assume that the signal processing means 2 outputs an acceleration signal on which power supply hum or the like is superimposed.

The short period signal in this signal is the power supply frequency, that is, 5
Let's assume it's a 0Hz hum. This signal is first input to the differentiating circuit 4, where the signal is differentiated, resulting in a waveform as shown in FIG. 2(B). This differentiated output is inputted to the first comparing means 5 and the second comparing means 6, respectively, and depending on whether the sign of the signal is plus or minus, if the sign is positive, the output from the first comparing means 5 is inputted to the first comparing means 5 and the second comparing means 6. , if the sign is negative, the second comparing means 6 outputs "H° times signal" as shown in FIG. 2 (D) and (E). These signals are outputted from the fifth comparing means 7 The signals are input to the sixth comparison means 8.The fifth and sixth comparison means 7 and 8 input the mechanical noise of the camera or the power supply frequency of 50Hz, which is not directly related to the camera shake signal, as described above. A comparison is made to remove hum, etc.

That is, as is clear from the signal waveforms in Figures 2 (D) and (E), only the hum component with a power supply frequency of 50 Hz is output here, and this is a comparative judgment that this is a signal that is not directly related to the camera shake signal. will be held.

Therefore, the output of the fifth and sixth comparison means 7.8 is as shown in FIG.
As shown in P) and (G), the "H2 signal is no longer output. Therefore, no signal is input to the first counting means 9 and the second counting means 10. Therefore, the average frequency calculating means 18 No signal is input from the 1.2 counting means 9.10.

On the other hand, the same signal input to the differentiating circuit 4, P,
Fil is also man-powered. In addition, P and Fl are composed of ordinary low-pass filters, so as shown in Fig. 2 (
From the acceleration signal as shown in A), a camera shake signal as indicated by the broken line g2 in FIG. 2(C) is output. Here, the dashed line g
The reason why the signal indicated by 2 is delayed in phase from the original signal indicated by the solid line p1 is due to the filter. Next, this is
The output signals of P and FIL are sent to the third comparison means 12.
and the fourth comparing means 13, respectively. This 3.4th comparison means 12.13, like the above-mentioned 1.2 comparison means 5.6, is separated from the respective comparison means 12.13 by sign status (FIG. 2 01).

The “H# times signal” shown in (1) is output. These signals may also have a frequency that is that of a camera shake signal, as described above, or may be caused by mechanical Φ noise. In order to determine whether the 7.8 comparison means 14.
15 respectively.

Here, since it is the camera shake signal itself, these 7.8
The output of comparison means 14.15 is as shown in Fig. 2 (J) and (
K), and the third and fourth counting means 16.
17. This 3rd degree 4th counting means 16.17
Now, in order to determine the frequency of the signal output from the comparison means 14.15 of the above 7.8 and use it as frequency data,
A clock signal (not shown) is used as the standard time, and the logical product of this clock pulse signal and the signals output from the seventh and eighth comparing means 14.15 is calculated. By counting this output, the frequency is determined from the number of pulses per unit time and converted into frequency data.
For example, this can be done using a D-type flip-flop circuit.

The frequency data outputted from the first to fourth counting means 9, 10, 16.17 is calculated by the average frequency calculating means 1 in FIG.
Sent to 8th. This calculation means 18 is composed of a counter unit that counts the number of times the frequency data is sent, and an integration unit that integrates the frequency data that is sent. Calculate the average frequency. The timing of calculating and updating this average frequency, that is, the timing at which calculation of the average frequency is started and the timing at which resetting is performed, is controlled by the control unit 21. This signal from the control section 21 is simultaneously sent to a cutoff frequency changing means 20, which will be described later, and an operation is performed to return the cutoff frequency of the low-pass filter in the signal processing means 2 to the setting in the reset state.

By the way, in the case of the acceleration signal waveform shown in FIG. 2(A), it is assumed that noise with a period of about 1/20 is mixed into the slow signal. In this case, the camera shake signal is clearly a gentler signal, and the average frequency calculation in this case is based on the output signal from the first counting means 9 shown in figure m1, that is, the signal shown in figure 2 (
F), the output signal from the second counting means 10 shown in FIG. 1, the waveform shown in FIG. 2(G), the output signal from the third counting means 16 shown in FIG. The waveform of FIG. 2 (J) and the fourth counting means 1 shown in FIG.
7, that is, the waveform shown in FIG. 2(K). However, the waveforms to be displayed in FIGS. 2(F) and (G) are divided into frequencies at FIGS. 2(D) and (E) from the differential waveform of FIG. 2(B), as described above. The first counting means 9 and the second counting means 10 determine that the signal is a high frequency component signal unrelated to the camera shake signal, and the signal is ignored and the output is zero. Therefore, the third counting means 16 and the fourth counting means 1
Only the waveforms shown in FIGS. 2(J) and 2(K), which are the output signals from 7, are valid in the average frequency calculation means 18, thereby making it possible to calculate only the original camera shake signal excluding high frequency components. .

Further, when the average frequency is calculated by the average frequency calculation means 18 shown in FIG. 1, the result is sent to the cutoff frequency change determination means 19. In this means 19, reference values of several frequencies are set, and the cutoff frequency of the low-pass filter in the signal processing means 2 is set to the optimum value at that time in accordance with the result of the average frequency calculation means 18. do.

By setting this cutoff frequency, the next cutoff frequency changing means 20 performs an operation to electrically switch the cutoff frequency of the low-pass medium filter in the signal processing means 2 using an analog switch or the like. The cutoff frequency will be changed. Therefore, when using a low-pass filter to remove high-frequency component signals unrelated to the blur signal, the phase delay of the blur signal itself that inevitably occurs can be minimized by setting the cutoff frequency of the low-pass filter to an optimal value. can be suppressed to a minimum.

As an example of the second input waveform, a signal with a period about 1/4 of the period of the gradual signal is symmetrically mixed in, superimposed on a gradual shaking signal as shown in FIG. 3(A). Assume that the acceleration signal shown in FIG. 1 is output from the signal processing means 2 in FIG. In this case as well, the average frequency calculating means 1
A signal will be output to 8. This process is the same as the first input waveform example, so a detailed explanation will be omitted, but only FIG. 3 will be explained below.

FIG. 3(A) shows an acceleration signal waveform showing an example of the second input waveform, and FIG. 3(B) shows the result when the acceleration signal of FIG. 3(A) is differentiated by the differentiator 4 of FIG. The signal waveform of Fig. 3 (C
) shows the acceleration signal in FIG. 3(A) in FIG.
, the signal waveform after removing high frequency components with Fil, FIG. 3(D) is the signal waveform shown in FIG. 1 of the signal in FIG. 3(B)
The signal waveform when only the positive component of the signal is extracted and converted into an "H" signal in the comparing means 5, FIG. 3(E) is the second comparison shown in FIG. 1 among the signals in FIG. 3(B). Means 6
The signal waveform in FIG. 3 (F) when only the negative component of the signal is extracted and converted into an "H" signal is obtained by inputting the output of the first comparing means 5 shown in FIG. 1 to the fifth comparing means 7. The signal waveform after removing the high frequency component as described above, the third
Figure (G) shows the signal waveform after the output of the second comparing means 6 shown in Figure 1 is input to the sixth comparing means 8 and high frequency components are removed as described above, and Figure 3 (H) shows the signal waveform. , Figure 3 (C) above
FIG. 3 (1) shows the signal waveform when the third comparing means 12 shown in FIG. The signal waveform when the signal indicated by the broken line g3 in C) is extracted by the fourth comparing means 13 shown in FIG. 1 and converted into an "H" signal, FIG. The output of the third comparison means 12 shown in the figure is input to the seventh comparison means 14, and the signal waveform after removing the high frequency component as described above, and the signal waveform in FIG. This is the signal waveform after the output of the fourth comparing means 13 shown in FIG.

In the case of the acceleration signal waveform shown in FIG. 3(A) above, a signal with a period approximately 1/4 of the period of the gradual signal is symmetrically mixed in, superimposed on the gradual signal. It is determined that the frequency component is the frequency of the co-educational blur signal. Therefore, the average frequency calculation in this case is performed by the average frequency calculation means 18 shown in FIG. ),
The average of the waveforms of (J) and (K) is calculated. Looking at these, the acceleration signal waveform shown in FIG. 3 (^) can be roughly divided into three frequency components, and the average of these frequencies is calculated by the average frequency calculation means 18 shown in FIG. , the optimum frequency at this time will be determined using the method described below.

Next, consider a case where an acceleration signal due to camera shake as shown in FIG. 4 is output from the signal output means 2 shown in FIG. 1. That is, suppose that the signal S1, which has a fairly slow frequency, suddenly changes to a signal S2, which is not due to mechanical noise of the camera, but has a fairly high frequency. In such a case,
In the above-described process of calculating the average frequency by the average frequency calculating means 18 shown in FIG. 1, the cutoff frequency of the low-pass φ filter in the signal processing means 2 cannot be made to follow the blur signal. . Therefore, a cutoff frequency forced change determination means 22 is separately provided, and an information signal for knowing the current cutoff frequency value is sent from the cutoff frequency determination means 19 to the counting means 3.4 shown in FIG. 16. The same information signals sent from 17 to the average frequency calculating means 18 are inputted to the forced change determining means 22, thereby setting the current cutoff frequency appropriately for the frequency of the current blur signal. Determine whether or not. These operations are performed by the cutoff frequency forced change determination means 22, and only if the cutoff frequency setting is not appropriate as a result of the determination, the cutoff frequency change means 20 changes the cutoff frequency to the optimal cutoff frequency at that time. Output a signal to change. The above operation is similar to the case where the signal shown in Fig. 4 is reversed, that is, when the signal S2 with a fairly high frequency changes to the signal S1 with a fairly slow frequency. .

By the way, the average frequency calculation means 18. The cutoff frequency change determination means 19 and the cutoff frequency forced change determination means 22 can perform the processing in an analog manner, but the processing takes time and the frequency of the camera shake signal is not always constant and changes quickly. Since it is not suitable for analysis, it is done using a microcomputer program built into the camera.

As described above, in this embodiment, first, the acceleration detection means for detecting camera shake signals built into the camera body detects high frequency components other than the original camera shake signals, such as camera mechanical noise, etc. Even if detected, it becomes possible to remove these high frequency components and extract only the original camera shake signal. Additionally, when removing high-frequency components, the average frequency of the camera shake signal is calculated to minimize the phase lag in the original camera shake signal, and the optimal cutoff frequency of the low-pass filter is set based on this. be able to. Even if the frequency of the camera shake signal changes rapidly, the cutoff frequency of the current low-pass competition filter is constantly compared with the frequency of the camera shake signal, and if it is judged to be inappropriate, it is forced to reach the optimal state. It becomes possible to change the removal rate of high frequency components.

[Effects of the Invention] As described above, according to the present invention, the mechanical noise of the camera that affects the camera shake signal of the camera body is removed by the low-pass filter, and the frequency of the camera shake signal is By making the cutoff frequency of the low-pass filter variable, it is possible to minimize the phase delay of the camera shake signal, while at the same time increasing the removal of high-frequency components.This allows the camera shake signal to be returned to its original form. It can be extracted as close as possible. Furthermore, even if the frequency of the camera shake signal changes suddenly, the effect on the performance can be reduced, and many other remarkable effects can be achieved.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a camera shake signal processing device showing an embodiment of the present invention. FIGS. 2 and 3 show the operation of each part in FIG. In the corresponding timing charts, Fig. 2 shows the case where noise with a period of about 1/20 is superimposed on the shake signal, and Fig. 3 shows the case where noise with a period of about 1/4 is superimposed on the shake signal. When the signals are mixed symmetrically,
Figure 4 is a waveform diagram showing a case where the shake signal suddenly changes from a signal with a fairly gentle frequency to a signal with a fairly high frequency, and Figure 5 is a cutoff frequency changing means in Figure 1 above. FIG. 1... Acceleration detection means (shake detection means)
2... Signal processing means 3... Integrating means (frequency identification means) 18
... Average frequency calculation means 19 ... Cutoff frequency change judgment means (cutoff frequency change means) 20 ... Cutoff frequency change means (cutoff frequency change means and cutoff frequency change means) Off-frequency forced change means) 22...Cut-off frequency forced change judgment means (
Cut-off frequency forced change means) 2nd 11〉 −−〉 m−〉 工” 工” 工” 工” 工” 工” 工 ” 工” (Spontaneous) No. 40 1. Indication of the incident 1999 patent application No. 260522 2, Name of the invention Camera shake signal processing device (037) Olympus Optical Industry Co., Ltd. “Detailed description of the invention in the specification” 6. Contents of amendment (1) From the bottom of the second page of the specification After the 4-truncated rCJ, “
3) Next to rD and CJ listed in the 1st line and 11th line of page 3 of the statement, add ``B.''. (4) In the same page 4, line 12, add ``during exposure'' after ``detection''. (5) In the same page, page 5, line 4, replace ``above'' with ``aforementioned.''
I will correct it. (6) After rP and FJ at the beginning of line 7 on page 9, ","
will join. (7) Add "," next to rL, P, and FJ on page 9, line 8.9. (8) Same page 10, line 2 and page 13, 9.10
Add "," next to rL, P, and FJ written in the line. (9) Each rL, P, as described in page 15, line 6 of the same
Add "," after FJ. (lO) rL, P, described in page 15, line 12 of the same
Add "," next to FJ.

Claims (2)

[Claims] (1) A shake detection means for detecting camera shake and outputting a shake signal; and at least a low-pass detection means for processing the shake signal.
a signal processing means including a filter; a frequency identification means for identifying the blur signal within the blur frequency range from the output of the signal processing means; and an average for calculating the average frequency of the blur signal identified by the frequency identification means. A blur signal processing device for a camera, comprising: a frequency calculating means; and a cutoff frequency changing means for changing the cutoff frequency of the low-pass filter based on the calculated average frequency. (2) The cutoff frequency changing means includes cutoff frequency forcing changing means for forcibly setting the cutoff frequency to a predetermined value when the amount of change in the identified blur signal exceeds a predetermined value. The camera shake signal processing device according to claim 1.