JPH06348284A - Active controller using adaptive digital filter - Google Patents

Abstract

(57) [Abstract] [Purpose] An object of the present invention is to provide an active control device which does not cause divergence in the adaptive coefficient used in the adaptive digital filter due to a sudden or temporary factor and is low in cost. [Structure] Control according to the stability of active control spectrum characteristics, which is particularly effective when the active control target sound does not change significantly in its spectral characteristics over a relatively long period of time and then becomes stable again if it changes in a relatively short period of time. In the method, the adaptive processing is not always performed by the adaptive algorithm with respect to the active control target sound, but during the stable period of the spectral characteristic of the active control target sound, the adaptive processing is stopped and the change of the spectral characteristic is monitored, During that time, the active control target sound is actively controlled as a fixed digital filter, and adaptive processing is continuously performed during the period when changes in the spectral characteristics are observed, and the error with the desired response is minimized by the active control processing until the stable period comes. Adaptively update the adaptive coefficient sequence such that

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal control circuit in an active silencer, a signal control circuit in active vibration control, etc.
Active control The present invention relates to a signal control circuit generally used.

[0002]

2. Description of the Related Art As disclosed in US Pat. No. 4,122,303, an active control device using an adaptive digital filter outputs an active control target sound (noise source) to an active control target sound input device (microphone). Etc.), and the signal sequence processed and output by the adaptive digital filter is output from the active control signal output device (speaker, etc.) to a predetermined external device (one-dimensional duct, etc.)
And output the response error sound of the sound wave interference with the active control target sound inside the external device with the response error sound difference input device (other microphone etc.), and the error with the desired response is minimized. As described above, the adaptive algorithm adaptively controls the coefficient sequence used in the adaptive digital filter.

Here, a conventional active silencer is shown in FIG.
Will be described with reference to. In the duct 2 for converting the active noise reduction target sound generated from the noise source 1 into the 0th mode, the noise generated from the noise source 1 is converted into an analog signal by the noise detection microphone 3, and after being appropriately amplified, a digital signal is obtained. Convert to. This digital signal is indicated by u. The signal u is input to the adaptive digital filter 6, and the active mute speaker 5 is driven by the processed and output signal string y. The vibration generated from the speaker 5 is combined with the noise in the duct 2. That is, the acoustic wave interference with the active control target sound from the noise source 1 is performed inside the duct 2, and as a result, the sound of the generated response error is acquired by the error detection microphone 4, and the error with the desired response is minimized. Then, the coefficient sequence A used in the adaptive digital filter 6 is controlled by the adaptive algorithm 8 by the output of the microphone 4. FIG. 5 is a flow chart showing the signal processing of this conventional example.

The fixed digital filter 7 has a transfer characteristic from the output section of the adaptive digital filter 6 to the input section of the adaptive algorithm 8 via the speaker 5, the duct 2, the microphone 4 etc. It is introduced to identify the transfer characteristic of the path from the output of the adaptive digital filter 6 to the input of the adaptive algorithm 8 by the adaptive processing of the algorithm 8. Here, the transfer characteristic is separately measured and stored in the storage unit 10 as a coefficient sequence B of the fixed digital filter 7.

[0005]

When the sound to be actively controlled is such that its spectral characteristics do not change significantly over a relatively long period of time and rarely changes in a relatively short period of time and then becomes stable again, When finely adaptive processing is continuously performed on the sound to be actively controlled, the adaptive coefficient used by the adaptive digital filter can obey these factors due to sudden or temporary factors such as disturbance or irregular device operation. There is a risk that learning may occur and that the adaptive coefficient may diverge. If the adaptive coefficient diverges, it is difficult for the adaptive algorithm of the prior art to recover normally by itself, and a separate means for recovering normality must be taken.

Further, in consideration of the cost reduction of the system, when the device for generating the active control target sound performs the arithmetic processing as its original function by the arithmetic processing device for the active control processing, the active control processing is performed. If the amount of processing required for execution per unit time is close to the maximum capacity of the arithmetic processing unit, there is extra capacity other than active control processing, and the original function of the device that generates active control target sound. It was difficult to control, and even if it did, it was necessary to keep it with only simple control. In order to solve this, if a higher speed arithmetic processing device is selected, there is a drawback that the price is increased. The present invention has been made in view of the above circumstances, and an object thereof is to provide an inexpensive active control device in which the adaptive coefficient used in an adaptive digital filter does not diverge due to a sudden or temporary factor.

[0007]

According to the present invention, the active control target sound is particularly effective when the active control target sound does not have a large change in its spectral characteristics for a relatively long period of time and is stable again when it changes in a relatively short period of time. It is a control method according to the stability of control spectrum characteristics, and does not constantly perform adaptive processing by an adaptive algorithm on the active control target sound, but stops the adaptive processing during the stable period of the spectral characteristics of the active control target sound. Then, the change in the spectral characteristics is monitored, while the fixed control target sound is actively controlled as a fixed digital filter,
The adaptive process is continuously performed during the period when the change in the spectral characteristic is recognized, and the adaptive coefficient sequence that adaptively minimizes the error from the desired response is adaptively updated by the active control process until the stable period comes.

[0008]

With this configuration, when the sound to be actively controlled does not have a large change in its spectral characteristics for a relatively long period of time and is stable again if it changes in a relatively short period of time, the spectral characteristics are compared. When it is in the stable state, the adaptive adaptive coefficient sequence is stopped to be updated during that period, and the state with the minimum error acquired up to that point is filtered as a fixed digital filter to detect the input signal. Perform active control. During that time, the influence of the disturbance disappears at all.

Further, in consideration of the cost reduction of the system, the processing contents of the arithmetic processing device as the original function of the device for generating the sound to be actively controlled is changed to the arithmetic processing device 1 for active control processing.
In the case of sharing between the two, at least in the stable period, the convolution calculation of the fixed digital filter for transfer characteristic identification from the speaker to the microphone and the calculation process of the adaptive algorithm for adaptively updating the coefficient sequence A are eliminated, and the spectrum of the active control target sound is eliminated. Since the process of monitoring changes in characteristics is relatively light, it is possible to sufficiently execute other jobs.
Therefore, due to the increase in the amount of processing per unit time due to the diversification of processing, according to the present invention, the need to select a higher speed arithmetic processing device is reduced.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS As an active control device using an adaptive digital filter, an embodiment of the present invention will be described below for a zero-order mode (one-dimensional) active noise reduction system. FIG. 1 shows the configuration of an active control device using an adaptive digital filter according to the present invention, and in particular, it is illustrated assuming a 0th-order mode (one-dimensional) active noise reduction system.

In the figure, the parts indicated by reference numerals 1 to 10 have been already described in the conventional example of FIG. 4 and therefore will be duplicated, but they will be explained again here. Reference numeral 1 is a noise source that continuously generates random noise. Reference numeral 2 denotes a duct as a predetermined external device, which is a duct configured so that a noise wave is established only in the 0th-order mode. Reference numeral 3 is a noise detection microphone for detecting noise. Reference numeral 4 denotes an error detection microphone for detecting a response error sound generated by synthesizing the noise and the output sound wave adaptively controlled by the adaptive digital filter 6.

Reference numeral 5 is an active noise reduction speaker for converting an output signal adaptively controlled by the adaptive digital filter 6 into a sound wave. Reference numeral 6 denotes an adaptive digital filter which processes a composite wave of an original noise wave and an output signal sound wave with respect to an input signal by a coefficient string adjusted so as to approximate a desired response by an adaptive algorithm and outputs the signal. . Reference numeral 7 denotes a coefficient string B for identifying the transfer characteristic of the path from the adaptive digital filter output section to the input section of the adaptive algorithm 8 via the speaker 5, the duct 2, the error detection microphone 4, etc. It is a fixed digital filter for considering the influence of the signal on the above-mentioned path to the adaptive processing. Reference numeral 8 applies the coefficient sequence of the adaptive digital filter 6 so that the difference between the error sound generated by the synthesis of the control sound wave processed by the adaptive digital filter 6 and the original noise wave and the desired response is minimized. It is an adaptive algorithm that adjusts dynamically. Reference numeral 9 is a memory of the adaptive coefficient sequence A used in the adaptive digital filter 6. This is adaptively updated by the adaptive algorithm 8 during the adaptive processing, and is replaced with the coefficient string that minimizes the response error during the non-adaptive processing. Reference numeral 10 is the coefficient sequence B described for the fixed digital filter 7.

Next, portions 11 to 14 are newly provided in this embodiment, and 11 of them is a memory for sequentially storing the adaptive update history of the coefficient sequence A for each sampling period. Reference numeral 12 is a memory for sequentially storing the adaptive update history of the moving average value E of the response error signal e0 from the error detection microphone 4 for each sampling cycle. Reference numeral 13 is a block for extracting the minimum value E (min) from the response error moving average history 12. Reference numeral 14 is a block for extracting the adaptive coefficient sequence H (best) corresponding to the minimum response error moving average value extracted by the minimum error extraction block 13.

FIG. 2 is a flow chart showing the flow of signal processing of the active control device using the adaptive digital filter of the present invention. First, the adaptive algorithm 8 is adaptively updated by the adaptive algorithm 8 according to the procedure described in the description of the conventional example until the moving average value of the response error e0 becomes smaller than a predetermined value, or even the adaptive coefficient convergence tendency. If it is known, it is continuously performed for a predetermined time or a predetermined number of times (# 5 to # 25). Meanwhile, both the adaptively updated adaptive coefficient sequence A and the corresponding moving average value of the response error signal are stored in the memories 11 and 12 as a history (# 40). It is desirable to take the moving average here by the number of sampling data corresponding to the wavelength of the noise to be silenced, but the accuracy is sufficient here so that the tendency of the change of the response error can be grasped.

When it is detected that the moving average value of the response error is smaller than a predetermined value for a predetermined period and the adaptation coefficient A is sufficiently converged (Yes in # 50), The processing of the adaptive algorithm 8 is temporarily stopped (# 53). Next, update history data 1 stored in memory
From 2, the coefficient sequence E (min) corresponding to the minimum data of the moving average of the response error is extracted (# 55). The coefficient sequence E (mi
n) is overwritten on the coefficient sequence A that has been used up to now by the adaptive digital filter 6 (# 60), and a so-called anti-phase waveform with respect to the input signal u of the active muffling target sound as a fixed digital filter by the coefficient sequence Is generated (# 70),
When sound wave interference is performed in the duct 2 (# 75), the response error sound is monitored by the error detection microphone 4, and when it is detected that the moving average value of the response error sound exceeds a predetermined value (# 80-
# 95), the adaptive algorithm 8 restarts the conventional adaptive coefficient updating process (# 100). At that time, the adaptive coefficient sequence A starts the adaptive processing with the coefficient sequence E (min) as an initial value.

The expressions relating to "sound" in the above description also apply to general mechanical "vibration". It should be additionally noted that the sound wave control of the present embodiment can be applied to the seismic wave control by using the noise detection microphone 3 as a vibration detection pickup or the like and the active noise reduction speaker as an active vibration control actuator or the like.

FIG. 3 shows the difference in control results when the same noise waveform is input as the sound to be muted in both the case of the prior art (FIG. 6) and the present invention. In FIGS. 3 and 6, active control is started at time T0, and both are converged at time T1 and the error is stable. In the present embodiment, which has recognized the stable state, the adaptive processing is interrupted and the fixed digital filter processing is switched to according to the coefficient sequence that causes the minimum error. Note that the conventional technique continues the adaptive processing as needed regardless of the stable state.

When the noise amplitude becomes large at T2, the conventional technique responds immediately, but this embodiment first recognizes that there is a change in the error, and then suspends the adaptive processing at time T2.
Restart from 3. Although there is some delay, it does not result in increased noise. In this embodiment, the stable period is recognized at time T4, and the adaptive processing is interrupted again. At time T5, the noise amplitude becomes small, and the respective processes are performed as described above. A state in which a so-called disturbance sound wave is mixed into the error detection microphone at time T8 will be described.

In the prior art, when the signal is large enough to cause an overflow in the AD converter, the DA converter, and the internal calculation, the control becomes uncontrollable and the state is reflected in the adaptive coefficient, and then the sound is muted. In an attempt to reduce the error caused by the failure, he / she outputs a large pulse noise, and this leads to a vicious circle. After that, the coefficient diverges and it becomes impossible to recover. However, in the case of the present invention, even if a momentary pulse noise such as T8 is mixed, it is not recognized as a state change and the adaptive processing interruption state is continued. Therefore, the error that occurs is only a momentary disturbance. And you don't make things worse for yourself.

[0020]

As described above, according to the present invention, the amount of processing of the arithmetic processing unit per unit time is kept to a minimum during the stable period in the convergence state of the adaptive coefficient sequence, while the divergence of the adaptive coefficient due to disturbance is suppressed. The active control process can be continuously performed without causing it.

Further, the random noise generated in response to the stepwise operation of the capacity switch such as a vacuum cleaner or a large machine tool as an example of the application and development of the present invention has a slight change in the noise spectrum depending on the load. , If the capacity switch is in the same position, there is a relatively stable period. It can be said that it is a stable period at least during operation under no load. In particular, during such a no-load operation, when a sound that was not input by the noise detection microphone is suddenly input from the error detection microphone, in the case of so-called disturbance, the present invention stops the adaptive coefficient update processing of the adaptive algorithm, It is possible to prevent the adaptive coefficient update process from being disturbed by a factor other than the sound to be actively muted.

[Brief description of drawings]

FIG. 1 is a diagram showing a signal processing configuration of an active control device using an adaptive digital filter embodying the present invention.

FIG. 2 is a flowchart showing a flow of signal processing of the active control device of FIG.

FIG. 3 is a diagram showing a waveform of a control result when the sound is input to the active control device of FIG. 1 as a sound to be muted.

FIG. 4 is a diagram showing a signal processing configuration of an active control device using a conventional adaptive digital filter.

FIG. 5 is a flowchart showing the flow of signal processing of a conventional active control device.

FIG. 6 is a diagram showing a control result as a waveform when the sound is input to a device of a conventional example as a sound to be muted.

[Explanation of symbols]

 1 Noise Source 2 Duct 3 Noise Detection Microphone 4 Error Detection Microphone 5 Active Silence Speaker 6 Adaptive Digital Filter 7 Fixed Digital Filter 8 Adaptive Algorithm 11 Coefficient Sequence A History Memory 12 Response Error Moving Average History Memory 13 Minimum Error Extraction Block 14 Coefficient sequence extraction block

Claims (2)

[Claims] 1. An error between the desired response and the response obtained by inputting the output signal to a predetermined external device while outputting the result of the convolution operation of the input signal sequence and the predetermined coefficient sequence as an output signal. In an active control device using an adaptive digital filter having an active control process for adjusting the coefficient sequence so as to minimize, the active control process is repeated at a specific cycle until the error becomes smaller than a predetermined value. During the adaptive process, the coefficient string for all taps in each specific cycle and the response error corresponding to the coefficient string are combined and stored in the memory, and adaptive update is performed when the error becomes smaller than a predetermined value. The processing is temporarily stopped, the minimum response error combination is extracted from the memory, and the coefficient string for all taps is transferred to the adaptive coefficient string of the current adaptive digital filter, An active control device using an adaptive digital filter, characterized in that the updating process is interrupted and the active control process is continued as a fixed digital filter. 2. An error between a desired response and a response obtained by inputting the output signal to a predetermined external device is constantly monitored while the active control process as the fixed digital filter is continued, and the predetermined error is monitored as needed. The active control device using the adaptive digital filter according to claim 1, wherein the adaptive update process is restarted when the value exceeding the value continues.