KR0118697Y1 - Noise detection device and noise reduction device - Google Patents

Abstract

The present invention is a device for automatically detecting the amount of noise contained in a signal and a noise reduction device for controlling the noise reduction operation based on the output of the detection device and suppressing the disturbance caused by performing noise reduction on a signal having a high S / N. Is to provide.

To this end, the present invention receives a signal input filter (1, 2, 3, 4, ...) to determine whether the noise signal, or how much the noise signal is included, and these filters (1 And, 2, 3, 4, ... are provided with an arithmetic unit 12 which comprehensively determines whether or not a noise signal is included or to what extent.

Description

Noise detector

1 is a block diagram showing the configuration of a noise detection apparatus according to a first embodiment of the present invention;

2 is a block diagram showing the configuration of a noise detection apparatus according to a second embodiment of the present invention;

3 is a block diagram showing the configuration of a noise detection apparatus according to a third embodiment of the present invention;

4 is a block diagram showing the configuration of a noise reduction device according to a fifth embodiment of the present invention;

5 is a block diagram showing the configuration of a noise reduction device according to a sixth embodiment of the present invention;

6 is a block diagram showing the configuration of a noise reduction device according to a seventh embodiment of the present invention;

7 is a block diagram showing the configuration of a noise reduction device according to an eighth embodiment of the present invention;

8 is a block diagram showing the configuration of a noise reduction device according to a ninth embodiment of the present invention;

9 is a block diagram showing the configuration of a noise reduction device according to a tenth embodiment of the present invention;

10 to 12 is a block diagram showing the configuration of a conventional noise reduction device,

13 is a waveform diagram showing a change in waveform when an input signal passes through a trade pass filter, a limiter, and a low pass filter in sequence;

14 is an explanatory diagram showing the arrangement of the contrast portions on the screen and signal lines constituting the screen;

15 is a block diagram showing the configuration of a noise reduction device according to a fourth embodiment of the present invention.

* Explanation of symbols for main parts of the drawings

1 ~ 4: filter 11,41: input terminal,

12,42: Output terminal 13,13a, 13b: Computing device

31 Inverter 32 AND circuit

33: OR circuit 43: noise detector

44,132 Subtractor 51,121 High pass filter

52,122,133: limiter 53,123: low pass filter

54,62,64,71a, 75b: GCA, 55,66,85: switch

61,74,84,131: Delay device 63,72,82: Adder

65,81,83: amplifier 95: noise reduction device

111,125,135: Comparator 112,126,136: Reference voltage source

124,134: Integrator 201 ~ 207: New Urun

211 ~ 222: synapse

[Industrial use]

The present invention relates to a device for detecting noise contained in a signal, and a device for performing noise reduction using the noise detection device.

[Traditional Technology and Problems]

Noise is likely to occur when a television, a VTR, an audio device, or the like performs signal processing, particularly when recording or reproducing a signal. In the past, it was very difficult to theoretically discriminate between the original signal and the noise signal. For this reason, there is no device for automatically detecting the presence or absence of a noise signal or the extent to which the noise signal is contained. Therefore, it was inevitable to judge the reproduced voice or image by human sense.

In addition, a device for processing video signals, audio signals, and the like is provided with a noise reduction device. A conventional noise reduction device used for a VTR will be described as an example. Noise cancellers include noise canceller, non-circulator, comb type filter and circulatory comb filter, and their block configurations are shown in FIGS. 10 to 12, respectively. Indicates.

The noise canceller has a configuration as shown in FIG. The signal input from the input terminal 41 passes through a high pass filter 51, a limiter 52, and a low pass filter 53. In the predetermined frequency band determined for the high pass filter 51 and the low pass filter 53, a component of the micro level determined by the limiter 52 is extracted, which is regarded as a noise signal and subtracted. After subtracting the signal from (44), it is output from the output terminal 42. In general, since the frequency band which is annoying to human being is around 1 MHz, it is performed to remove micro components in this band.

The non-circulating comb filter suppresses randomly generated noise signals using two correlated signals, and has a configuration as shown in FIG. Like the VTR signal, a signal correlated by any one of one horizontal period, one field period, or one frame period is localized by the delay device 61, and a signal input from the input terminal 41 is input to this delayed signal. The adder 63 adds the result. Then, the signal output from the adder 63 is amplified by the amplifier 65 1/2 times to return to the original signal level and output from the output terminal 42. In this way, by using two correlated signals, it is possible to amplify and suppress (1/2) ^0.5 the level of noise generated randomly without correlation.

The cyclic comb filter also suppresses the influence of the noise signal by using two correlated signals, and the signal input from the input terminal 41 is multiplied by 1-k (0k1) by the amplifier 71. In addition, another signal delayed by the above-described period by the delay circuit 74 is k times by the amplifier 75. The signal output from the amplifier 75 and the signal output from the amplifier 71 are added by the adder 72 and output from the output terminal 42.

However, even when any noise reduction device is used, there is a problem that the image quality of a reproduced image is deteriorated. For example, in the noise canceller device, not only the noise but also the original signal component is contained in the predetermined frequency band extracted by the high pass filter 51, the limiter 52 and the low pass filter 53. For this reason, the original signal components as well as the noise signal are subtracted from the signal, resulting in a decrease in resolution.

In addition, in the case where a non-circulating filter or a cyclic filter is used, the same operation is performed for two uncorrelated signals, so that an afterimage occurs in the vertical direction of the screen.

For signals with low S / N ratios, the use of these filters is effective in improving the S / N ratio, but conventionally, the presence or absence of a noise signal included in the signal as described above There was no device to automatically detect the degree. For this reason, noise reduction is also performed for signals having a high S / N ratio and not requiring much noise reduction, thereby causing a problem of degradation in resolution.

As described above, conventionally, there is no device for automatically detecting the presence or absence of a noise signal included in a signal and the degree of the inclusion of a noise signal, which is inevitably determined by the human eye or ear.

In addition, in the conventional noise reduction device, since there is no device that automatically detects the presence or absence of a noise signal, noise reduction is performed on all signals regardless of the S / N ratio, resulting in a deterioration of image quality and sound quality. Was done.

[Purpose of designation]

The present invention has been made in view of the above circumstances, and an object thereof is to provide a noise detection device capable of automatically detecting the presence or absence of a noise signal and a noise reduction device using the noise detection device.

[Composition of design]

In order to achieve the above object, the noise detection apparatus of the present invention includes at least two input signals that determine whether a noise signal is included in the signal, or how much of the noise signal is included in the signal. Based on the filter and the results determined by the filter, the filter determines whether the noise signal is included in the signal or how much the noise signal is included in the signal, and outputs the determined result. It is characterized by including a computing device.

Here, at least one of the above filters may be regarded as the noise signal to determine components below a predetermined level within a predetermined band.

At least one of the filters may delay another signal that has a correlation with the signal, and may determine that a component having no correlation between the delayed other signal and the signal is regarded as the noise signal, or the correlation It may be determined by considering that the noise signal is less than or equal to a predetermined level among the missing components.

When the signal is a video signal, a filter for determining whether the brightness of the screen indicated by the video signal is equal to or less than a predetermined level may be further included.

The computing device may be a logic circuit or a neural network system.

In addition, the noise reduction device of the present invention includes a noise canceller device which receives a signal and outputs a component having a predetermined level or less within a predetermined frequency band of the signal, a noise detection device having the above-described configuration, and And control means for controlling the characteristics of the noise canceller device based on the result determined by the computing device equipped with the noise detector.

Here, the noise canceller is operated when the computing device determines that the signal contains the noise signal instead of the control means, or when the signal includes more than the predetermined level of the noise signal. In the case where it is determined that the noise signal is not included in the signal or when it is determined that the noise signal is not included in the signal more than a predetermined level, the noise canceller device may be provided with switching means for not operating the noise canceller device. Instead of the noise canceller body, a comb filter may be provided.

[Action]

According to the present invention configured as described above, at least two filters determine whether or not a noise signal is included in the input signal or how much the noise signal is included, and based on the respective determination results, By comprehensively determining, it is possible to automatically and accurately detect the noise signal.

When the filter detects a noise signal, the criteria for judgment vary depending on which component included in the signal is to be regarded as a noise signal. Therefore, if necessary, a component having a level below a predetermined level or another signal having a correlation within a predetermined number of abacus bands can be regarded as a noise signal, for example.

If the signal is a video signal, and further includes a filter for comparing the brightness of the screen with a predetermined level, it is possible to determine whether the noise signal is in a prominent situation.

When the number of filters is small, the circuit size can be reduced by using a logic circuit as the computing device. On the contrary, when the number of filters is large, the size using the neural network system can be reduced.

By controlling the characteristics of the noise canceller device using the control means based on the determination result of the noise detection device, it is possible to perform the appropriate noise canceller operation according to the presence or absence of the noise signal or the degree to which the noise signal is included.

If it is determined that the noise signal does not exist or the noise signal is not included in the noise detection device, the noise canceller device is not operated.

The same applies if a comb filter is used instead of the noise canceller device.

EXAMPLE

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

1 shows a block configuration of a noise detection apparatus according to a first embodiment of the present invention, in which signals input from the input terminal 11 are input to various filters 1, 2, 3,... The filters (1, 2, 3, ...) have a criterion for judging the presence or absence of a noise signal included in the signal, or whether the noise signal is prominent in the human eye as described below. Each one is different. Therefore, the result of determination based on the criteria set in advance is output from each filter 1, 2, 3, ..., and the determination result of each filter is input to the calculating device 13. The arithmetic unit 13 receives these plurality of determination results and comprehensively judges the presence or absence of a noise signal and outputs the final determination result from the output terminal 12.

As described above, according to the first embodiment, it is judged whether or not the presence of the noise signal or the situation where the noise signal is prominent on the basis of the respective determination criteria in the various filters, and the calculation is performed based on these determination results. It is determined comprehensively in the device 13. Therefore, the presence or absence of noise or the degree to which the noise is included can be automatically and accurately determined.

2 shows the configuration of the second embodiment in which the first embodiment is more specific, and the filter 1 determines whether or not the noise is in a prominent situation based on the brightness of the screen. This is based on the fact that even if the noise level is about the same, it becomes noticeable to the human eye as the screen darkens. In the comparator 111, the level of the signal input from the input terminal 11 and the reference voltage VA applied from the reference voltage source 112 are compared. If the level of the signal is lower than or equal to the reference voltage VA, the captured signal is likely to stand out, and a signal of logic [1] level is outputted to the node 21. In other cases, the signal of logic [0] level The signal will be output.

The filter 2 regards a small signal in the frequency band near 1 MHz in which the noise signal stands out as a noise signal. The high pass filter 121, the limiter 122, the low pass filter 123, the integrator 124, and the comparator 125 are connected in series to determine whether or not the degree to which the captured signal is contained is a predetermined level or more. Is connected. The high-pass filter 121, the low-pass filter 123, and the limiter 122 extract the minute signal components included in the predetermined frequency band, which is regarded as a noise signal for the integrator 124 for a predetermined time. Integrate. The integrated value is compared with the reference value applied from the reference voltage source 126 by the comparator 125 and the comparison result is output to the node 22. Here, a signal of logic " 1 " level is output when the integral value is equal to or greater than the reference value, that is, when it is determined that the noise signal is included in the predetermined level or more.

The filter 3 compares two correlated signals and judges the presence or absence of the captured signal. Delay device 131 delays one of the correlated signals. This delay time corresponds to any one time of one horizontal period, one field period, or one frame period. By subtracting this delayed signal from the other non-delayed signal with respect to the subtractor 132, an uncorrelated signal is output to the limter 133. The original signal component is also included in this signal, and the level of such components is often above a predetermined level. Thus, the minute components passing through the limiter 133 are regarded as noise signals and output to the integrator 134. In the integrator 134, the input signal component is integrated for a predetermined time and output to the comparator 135. The comparator 135 compares the output integral value with the reference value applied by the reference voltage source 136 and outputs the result to the node 23. Also in this filter 3, similarly to the filter 2, the logic [1] level is output when the integral value is more than the reference value.

The outputs from the filters 1 to 3 are input to the arithmetic unit 13a. The arithmetic unit 13a includes an inverter 31, an AND circuit 32, and an OR circuit 33. The three input terminals of the ADN circuit 32 have a node 22, a node 23, and an inverter ( The nodes 21 are connected to each other via 31. The two input terminals of the OR circuit 33 are connected to the output terminals of the node 21 and the AND circuit 32, respectively. The output terminal is connected to the output terminal 12.

Table 1 below shows the relationship between the input and output signal levels of the computing device 13a.

The relationship between these input and output levels is as follows.

When a signal of logic " 1 " level is outputted from the filter 1, the screen is darker than a predetermined level, so that the noise signal stands out.

Therefore, when at least one of the filter 2 and the filter 3 judges that a noise signal exists, and outputs a signal of logic "1" level, the signal of logic "1" level is output.

When a signal of logic " 0 " level is output from the filter 1, there is a situation where the noise signal is not prominent. Therefore, only when the filters 2 and 3 output the signals of the logic "1" level, the signals of the logic "1" level are output.

As described above, according to the second embodiment, it is determined whether or not the noise signal is prominent by the filter 1, and the filter 2 determines whether or not a minute component is contained in a predetermined frequency band or more. The presence or absence of a noise signal is determined by the presence of a noise signal, and whether or not a random noise signal exists by a predetermined level or more is determined by the two signals correlated by the filter 3. And the arithmetic unit 13a judges the presence or absence of a noise signal comprehensively based on each determination result of the filters 1-3. In this way, by comprehensively determining using the determination results of a plurality of filters having different criteria, it is possible to accurately and automatically determine the presence or absence of a noise signal.

Next, a third embodiment of the present invention will be described with reference to FIG.

In this embodiment, although the filters 1 to 3 are the same as in the second embodiment, the arrangement of the arithmetic unit 13b is different. That is, while the computing device 13a in the second embodiment is constituted by a logic circuit, the computing device 13b in the third embodiment is constituted by a neural network system of three layers.

Neural network systems have been put to practical use in recent years, and have realized the function of nerve cells in the brain by using a number of simple computing elements. Then, an input signal is applied to the neural network, and the input signal is trained to obtain a desired output signal. The input / output signal is repeatedly applied until the desired output signal or a result close to the desired output signal is obtained to determine the weight count group. As such an algorithm, back propation is the most common.

This arithmetic unit 13b is provided with the synapses 211-222 provided between the neurons 201-207 and the signal line which connects each neuron 201-207. The neurons 201 to 203 constitute an input layer, the neurons 204 to 206 form an intermediate layer, and the neurons 207 constitute an output layer. The neurons 201 to 207 each add all the input data, and when the added value exceeds the threshold, for example, the operator outputs a logic "1". The synapses 211 to 222 change the strength of coupling between the neurons 201 to 207 and multiply the weight coefficient by the data flowing through the signal line. Backpropagation learning is performed so that the weight coefficient of each of the synapses 211 to 222 is optimal. Specifically, a video signal is input to the input terminal 11 of this embodiment. The filter 1 determines whether this signal is in a situation where the noise signal is prominent, and the filters 2 and 3 judge whether the noise signal is included in a predetermined level or more. The respective determination results of the filters 1 to 3 are input to the computing device 13b so that the output results are output from the output terminal 12. This output result is finally determined as the backpropagation is performed so that the weight coefficient of each synapse (211 ~ 222) is changed to approach the desired value.

Neural network system is effective in making a comprehensive judgment by processing a lot of data at the same time. Therefore, when the number of filters is large, the computing device 13b using the neural network system in the second embodiment is more suitable than the computing device 13a composed of the same logic circuit as the first embodiment. For example, it is also effective to provide a plurality of filters 1 having different reference potentials of the comparator 111. Similarly, a filter having a different frequency band determined by the high pass filter 121 and the low pass filter 123, the integral time of the integrator 124, the reference potential of the fertilizer 125, or the reference potential of the limiter 124. A plurality of (2) s may be provided, and a plurality of criteria for determining noise level may be provided to apply a plurality of input signals to the computing device.

Moreover, in the case of using a neural network system, there is an advantage that a desired output signal is obtained even when an unknown input signal is applied. In the process of backpropagation learning, a plurality of input / output signals are applied to determine the weight coefficient. Even after an unknown signal is input after this learning, a windy signal is output by the learning effect.

Here, in the second embodiment, the arithmetic unit 13b composed of the back-progression neural network system is used, but other systems may be used. For example, a system for learning by a hop field network system or a vector quantization method, a Boltzmann machine type system, an RCE network system, a neocognetron type system The same effect can be obtained even if it is used.

Further, the computing device is not limited to the ones shown in the first and second embodiments, and may be used to determine whether or not a noise signal is present comprehensively by using data output from a plurality of filters. For example, a microcomputer or the like that can be programmed by a designer may be used so as to obtain a desired input / output relationship.

The signals input and output to the arithmetic units 13a and 13b of the second and third embodiments adopt the form of digital signals, but the analog signals may be input or output. In this case, there is no noise signal. The degree of noise signal is displayed steplessly.

Next, a noise reduction apparatus according to a fourth embodiment of the present invention will be described. FIG. 15 shows the block configuration of the present embodiment, and between the input terminal 41 and the output terminal 42, the noise reduction device 95 and the noise reduction device 95 as shown in FIGS. Control means (96) for controlling the operation of the < RTI ID = 0.0 >) < / RTI > The noise detector 43 can use the noise detector 43 according to the first to third embodiments. The control means 96 controls the operation of the noise reduction device 95 based on the result of determining the presence or absence of the noise signal detected by the noise detection device 43 or the degree of the noise signal.

As described above, the conventional noise reduction device 95 performs noise reduction on all signals regardless of the s / n ratio of the input signal. As a result, when the S / N ratio of the signal is high, the obstacle of deterioration of image quality is caused. Therefore, the noise reduction device of this fourth embodiment detects the presence or absence of the captured signal contained in the input signal or the degree to which the noise signal is contained by the noise detection device 43, and accordingly the noise reduction device 95 By switching the operation, non-operation, or changing the strength of the noise reduction effect, the noise reduction effect is weakened when the S / N ratio of the input signal is high, so that the disturbance of signal degradation can be minimized.

In the following, the noise reduction device according to the fourth embodiment is further specified. 4 shows a block configuration of the apparatus according to the fifth embodiment of the present invention. The fifth embodiment shows a high pass filter 51, a limiter 52, a low pass filter 53, In addition to the subtractor 44, a noise detector 43 in the fourth embodiment and a gain control amplifier 54 (hereinafter referred to as GCA) are further provided. A GCA 54 is connected to the rear end of the low pass filter 53, and its output end is connected to the input terminal of the negative side of the subtractor 44. The output terminal of the noise detector 43 is connected to the control terminal of the GCA 54.

When a signal is input from the input terminal 41 and passes through the high pass filter 51, the repeater 52, and the low pass filter 53 in order, the signal waveform changes sequentially. In FIG. 14, it is assumed that the central part of the screen 90 has a high-brightness part 92 and a black and low-brightness part 91 around it. Among the signals constituting such a screen 90, the illustrated signal 93 has a waveform as shown in Fig. 13A. Here, the signals in the periods P1 and P3 are synchronous signals, and the video signals constituting the screen appear only during the period P2.

In the step input to the input terminal 41, the high frequency and low frequency noise components are included as shown in FIG. Passing through the high pass filter 51, the low frequency components are removed as shown in FIG. Furthermore, when passing through the limiter 52, the level above the predetermined level is removed as shown in FIG. 13 (c), and after passing through the low pass filter 53, low frequency components are removed as shown in FIG. 13 (d). do. Although the signal passing through the low pass filter 53 is composed of a noise signal that is most prominent on the screen, the original signal is also included to some extent.

Thus, the noise detection device 43 is used to determine the presence or absence of noise or the degree to which the noise is included, and change the gain A of the GCA 54 according to the determination result. When the amount of noise with a low S / N ratio of the signal is large, the gain A is set to a large value close to one.

Accordingly, in the subtractor 44, the gain A is multiplied by the signal output from the low pass filter 53, and then subtracted from the input signal to obtain a stronger noise reduction effect. When the signal S / N is high, the gain A is set small so that the signal output from the low pass filter 53 is not subtracted from the input signal too much.

5 shows a noise reduction device according to a sixth embodiment of the present invention. In this embodiment, the GCA 54 in the fifth embodiment is replaced by a switch 55. As shown in FIG. This switch 55 is turned on when the noise detector 43 determines that the amount of noise is large. As a result, the noise signal is removed from the signal. On the other hand, when the noise detection device 43 determines that the amount of noise is small, the switch 55 is off-door and noise reduction is not performed. As a result, obstacles caused by noise reduction are suppressed.

6 shows a configuration of a noise reduction device according to a seventh embodiment of the present invention.

In this embodiment, a non-circulating comb filter composed of a delay device 61 and an addition time 63 is used. The comb filter uses two correlated signals to compress a random noise signal without correlation, one horizontal period and one field period. Alternatively, the delay signal is used for one period of one frame period.

The rear end of the delay device 61 is provided with a GCA 62, and the rear end of the adder 63 is provided with a GCA 64. The gain B of the GCA 62 and the gain C of the GCA 64 are respectively controlled according to the amount of noise detected by the noise detector 43. Here, the gains B and C are controlled so that the level of the signal output from the output terminal 42 becomes constant, that is, the relationship (1 + B) C = constant holds.

The real signal input from the input terminal 41 and the signal delayed by the delay device 61 and multiplied by the gain B are added by the adder 63. The gain C is multiplied by this added value and output from the output terminal 42. In this case, when the noise detector 43 determines that a large amount of noise signal is included, the gain B is set to a large value close to 1, and, on the contrary, is determined to be small when it is determined that it is not included much.

In the eighth embodiment shown in FIG. 7, the non-circulating comb filter is used in the same manner as the seventh embodiment, but the switch 66 is used instead of the GCAs 62 and 64 to reduce the noise on and off of the filter. The control is different depending on.

When the noise detector 43 determines that the amount of noise is less than the predetermined level, the switch 66 is turned off and directly output without applying noise reduction to the input signal. When it is determined that the amount of noise is large, the switch 66 is turned on to reduce noise as follows. The signal delayed by the delay circuit 61 and the non-delayed signal are added by the adder 63 at the same level. By using the amplifier 65 whose gain is 1/2 with respect to the added signal, the level of the signal is returned to the normal level and the level of the irrelevant noise signal is reduced (1/2). To reduce.

Next, FIG. 8 shows a configuration of a ninth embodiment of the present invention using a circulation comb filter. In this embodiment, the circulation coefficient K representing the magnitude of the effect of the circulation comb filter is controlled in accordance with the amount of noise. Compared with the conventional circulation comb filter shown in FIG. 12, the amplifiers 71 and 75 are replaced with the GCAs 71a and 75a.

The noise detector 43 detects the amount of noise included in the input signal, and changes the gain K of the GCA 75a and the gain 1-K of the GCA 71a according to the detection result. . If it is determined that the amount of noise is large, the gain K is set to a large value close to 1, and if it is determined to be small, it is set to a smaller value. The signal multiplied by the gain 1-K is delayed by the delay device 74 so that the signal multiplied by the gain K is added by the adder 72 and output.

The tenth embodiment shown in FIG. 9 differs from the ninth embodiment in that the operation of the forward-shaped cone filter is turned on and off in accordance with the amount of noise using the switch 85. When the amount of noise is detected by the noise detector 43 and the detected amount of noise is small, the switch 85 is turned off and the input signal is output as it is. However, when the amount of noise is large, the operation is the same as that of the apparatus of FIG. In other words, the gain 1-K is multiplied by the input signal, which is delayed and the signal multiplied by the gain K is added to be output.

As described above, all of the noise suppressors according to the fourth to tenth embodiments of the present invention control the operation of various filters in accordance with the output of the noise detector 43. Here, the connection relationship of GCA or arrangement and number of switches in each illustrated embodiment is not limited to this, and controls the effect of noise reduction or turns on or off the noise reduction operation according to the amount of noise. Is good. In addition, the device controlled or turned on and off by the noise detector 43 is not limited to the noise reduction device according to the embodiment, but can be used for all noise reduction devices or other systems requiring noise protection.

[Effect of design]

As described above, according to the noise detecting apparatus of the present invention, at least two filters determine whether or not a noise signal is included in the input signal, and based on the result determined by each filter. Therefore, since it is determined comprehensively by the computing device, it is possible to accurately and automatically detect noise.

Further, according to the noise reduction device of the present invention, since the characteristics of the noise reduction operation are controlled based on the result determined by the noise detection device, it is possible to suppress the disturbance caused by the noise reduction.

Claims (6)

At least two filters adapted to input a signal to determine whether a noise signal is included in the input signal; An arithmetic unit adapted to comprehensively determine whether the noise signal is included in the input signal based on the results determined by the filters, and output the determined result; And said computing device comprises a neural network system having a plurality of neurons and a plurality of synapses for setting the strength of coupling each neuron. At least two filters adapted to input a signal to determine whether a noise signal is included in the input signal; An arithmetic unit adapted to comprehensively determine whether the noise signal is included in the input signal based on the results determined by the filters, and output the determined result; A plurality of first neurons adapted to receive the results determined by the filter, respectively, and to output a first comparison result in comparison with a predetermined criterion; A plurality of first synapses adopted for setting the strength by which the first neuron and the plurality of second neurons are combined by weighting; A plurality of second synapses adopted for setting the strength for coupling the second neuron and the third neuron by abrupt attachment; The plurality of second neurons are output from the first neurons and are respectively adapted to receive the first comparison results weighted by the first synapse and to output second comparison results in comparison with predetermined criteria. And whether the noise signal is included in the input signal by receiving the second comparison result, which is output from the second neuron and the second comparison result weighted by the second synapse, and compared with a predetermined criterion. Noise detection device, characterized in that it is adopted to output a determination result by determining comprehensively. At least two filters adapted to input a signal to determine how much a noise signal is included in the input signal; An arithmetic unit adapted to comprehensively determine how much the noise signal is included in the input signal based on the respective determination results of the filters, and output the determination result; And the computing device comprises a neural network system having a plurality of neurons and a plurality of synapses for setting the strengths of coupling each neuron. At least two filters adapted to input a signal to determine how much a noise signal is contained in the input signal; An arithmetic unit adapted to comprehensively determine how much the noise signal is included in the input signal based on the respective determination results of the filters, and output the determination result; A plurality of first neurons adapted to receive the results determined by the filter, respectively, and to output a first comparison result in comparison with a predetermined criterion; A plurality of first synapses adopted for setting the strength for coupling the first neuron and the second neuron by weighting; A plurality of second synapses adopted for setting by carrying out a weighting rule for joining the second neuron and the third neuron; The second neuron is output from the first neuron and is weighted by the first synapse and is adapted to receive the first comparison result and output a second comparison result in comparison with a predetermined criterion, respectively. The noise signal is included in the input signal based on the output of the third neuron from the second neuron and the second comparison result weighted by the second synapse. Noise detection device, characterized in that it is adopted to output a judgment result by determining comprehensively. At least one filter adapted to input the signal to determine whether the noise signal is included in the input signal, and the signal to be input to determine how much the noise signal is included in the input signal At least one filter; An arithmetic unit adapted to comprehensively determine how much the noise signal is included in the input signal based on the results determined by each of the filters, and output the determined result; And said computing device comprises a plurality of neurons and a neural network system having a plurality of synapses for setting the strengths that couple each neuron. At least one filter adapted to input the signal to determine whether the noise signal is included in the input signal, and the signal to be input to determine how much the noise signal is included in the input signal At least one filter; An arithmetic unit adapted to comprehensively determine how much the noise signal is included in the input signal based on the respective determination results of the filters, and output the determination result; A plurality of first neurons adapted to receive the results determined by the filter, respectively, and to output a first comparison result in comparison with a predetermined criterion; A plurality of first synapses adopted for setting the strength for coupling the first neuron and the second neuron by weighting; A plurality of second synapses adopted for setting the strength for coupling the second neuron and the third neuron by weighting; The second neuron is output from the first neuron and is adapted to receive the first comparison result weighted by the first synapse, respectively, and output a second comparison result in comparison with a predetermined criterion. A third neuron is output from the second neuron and the second comparison result is weighted by the second synapse, respectively, and the fertilizer is compared with a predetermined criterion to determine how much of the noise signal is included in the input signal. Noise detection device characterized in that it is adopted to output a judgment result by comprehensively determining whether or not.