JPH09139676A - Data conversion method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain higher reliability even in the case of expansion without much change from original data. SOLUTION: A correction signal generating section 5b of the converter uses an original signal V0 , an inverse conversion signal VI0 and a permissible error (e) to compare the original signal and n-th data of the inverse conversion signal. When the absolute value of the error is not in excess of the permissible error (e), the n-th data of the correction signal are set to 0, and when the absolute value of the error is in excess of the permissible error (e), the n-th data of the correction signal are replaced with n-th data of the original signal. The correction signal generating section 5b generates correction signals VX0 :VX0 (n) consisting of N-sets of data by conducting the processing above from n=1 to n=N.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data converter for compressing / expanding image data, audio data, process data, and measurement data such as pressure and voltage.

[0002]

2. Description of the Related Art Irreversible data compression techniques are roughly classified into a direct compression method for directly deleting redundant data and a conversion compression method for extracting features by orthogonal transformation. Of these, most of the data compression methods for which the allowable error can be set are classified into the direct compression method. Among the direct compression methods, the data compression methods for which the allowable error can be set include the zero-order prediction type compression method, the first-order prediction type compression method, the zero-order interpolation type compression method, and the first-order interpolation type compression method. There is. All of these methods are based on linear approximation of the entire data.

Among the above-mentioned methods, the linear interpolation type compression method has the highest compression rate. The outline of the linear interpolation type compression method will be described below. First, the data at a certain time tn is set as Xn, and the starting point X0 is stored as a required point.
Next, draw two straight lines from the starting point X0 to points above and below the next point X1 by the allowable range e, and draw these U1 and L
Let it be 1. Similarly, from the starting point X0 to the second next point X
Two straight lines U2, L2 to the upper and lower points by the allowable range e from 2
pull.

Here, if there is no overlapping portion between the area sandwiched by the straight lines U2 and L2 and the area sandwiched by the straight lines U1 and L1, the point X1 is saved as necessary data, and this X1 is stored.
With the new starting point, do the same as above. on the other hand,
If there is an overlapping part, the point X1 is regarded as extra data and deleted, and this time, two straight lines are drawn from the starting point X0 to a point above and below the allowable range e from the point X3 by U.
3, L3.

When there is a portion where the region sandwiched by the straight lines U3 and L3 and the region sandwiched between the straight lines U2 and L2 and the straight lines U1 and L1 overlap, the point X2 is regarded as extra data and deleted. Stores the point X2 as necessary data. Similarly to the above, two straight lines U are provided from the starting point Xm to points above and below the point Xn by the allowable range e.
n and Ln are subtracted, and the area between these two straight lines and k
= M + 1 to n−1, two straight lines Uk and L
If there is a portion that overlaps the area sandwiched by k, the point Xn-
1 is regarded as extra data and deleted.

In addition to the above-mentioned method, there is a data compression technique using wavelet transform. The wavelet transform is classified into a continuous wavelet transform and a discrete wavelet transform according to its property. The discrete wavelet transform is classified into an orthogonal wavelet transform and a non-orthogonal wavelet transform depending on whether or not the inverse transform is uniquely determined.

Since the wavelet transform used for data compression is an orthogonal wavelet transform, the orthogonal wavelet transform will be described below. The signal on the time domain is V ₀ : V ₀ (n) {n = 1, 2, ..., N}. The orthogonal wavelet forward transform is expressed by the following equation 1.

[0008]

(Equation 1)

According to this conversion formula, from the signal V ₀ to the signal V ₁ : V ₁ (n) {n = 1, 2, ... N / 2} and W ₁ : W
₁ (n) {n = 1,2, ... N / 2} is obtained, and similarly,
From V _j : V _j (n) {n = 1, 2, ~ N / 2 ^j }, V ^j
_{j + 1} : V _{j + 1} (n) {n = 1, 2, ~ N / 2 ^{j + 1} } and W
_{j + 1} : W _{j + 1} (n) {n = 1, 2, ... N / 2 ^{j + 1} } is obtained. In addition, the orthogonal wavelet inverse transformation is given by
It is represented by

[0010]

(Equation 2)

Therefore, in order to obtain the original signal V ₀ in the time domain by the inverse transform, the signals V _j and W _{1 to} W _j {j is a natural number} obtained by the wavelet forward transform are required. When the data number N of the signal V _{0 in} the time domain is divisible by 2 ^j , the sum of the data numbers of V _j and W _{1 to} W _j is N,
It becomes equal to V ₀ .

From the above, the orthogonal wavelet transform can be said to be a transform that decomposes the signal V ₀ in the time domain into a plurality of signals V _{j in} the time frequency domain and W _{1 to} W _j . Further, since V _j is a signal which is obtained by passing the original signal through a low pass filter, it is called a smoothed signal, and W _{1 to} W _j are signals which are passed by a band pass filter, and thus are detailed signals. Called. Data compression using wavelet conversion is performed by the following procedure.

The original data is converted into the signal V ₀ : V ₀ (n) {n
= 1, 2, ..., N}, the orthogonal wavelet forward conversion is repeated j times for V ₀ to obtain the smoothed signal V _j and a plurality of detailed signals W _{1 to} W _j . Quantization is performed on each of the smoothed signal V _j and the detailed signals W _{1 to} W _j , and data with small amplitude in each signal is deleted. A combination of the quantized smoothed signal V _j and the quantized detailed signals W _{1 to} W _j is used as compressed data. Decompression of data is performed by subjecting compressed data to inverse inverse wavelet transform.

[0014]

In the data compression / decompression technique capable of setting the allowable error by the above-mentioned conventional method, the target signal is approximated by a straight line. The discontinuity which was not in the signal of and occurs. This problem is noticeable when the target signal is smooth. Therefore, the signal expanded by this method has different continuity and frequency characteristics from those of the original signal, and could not be used for abnormality diagnosis to see the continuity of the signal or frequency analysis of the signal. .

On the other hand, in the conventional data compression using the wavelet transform, the compressed / decompressed data is
The discontinuity that does not exist in the original signal does not occur. However, since this wavelet transformation is expanded only by using a certain function and the data itself cannot be modified, the data compression cannot set the allowable error for the original signal. For this reason, there is a problem in that it is difficult to perform signal processing using a threshold value because the decompressed data points are not reliable.

The present invention has been made to solve the above problems, and when the compressed data is decompressed by the data compression by the wavelet transform, the original data is not much different from the original data. The purpose is to obtain higher reliability.

[0017]

According to the data conversion method of the present invention, the original signal is wavelet-converted with a desired number of expansions to generate converted data, and the converted data is inversely wavelet-converted to obtain the original signal. The converted data is modified so that the difference is within a desired allowable error range, and the original signal is compressed to generate compressed data. Therefore, the decompressed data obtained by decompressing the obtained compressed data is within the allowable error set from the original signal. Also, if the difference between the original signal and the value obtained by inverse wavelet conversion of the smoothed signal in the converted data exceeds the desired tolerance, a corrected signal that is restored to the original signal is generated instead of the smoothed signal. The data consisting of the number of expansions, the smoothed signal and the modified signal is generated as compressed data obtained by compressing the original signal. In this case, the obtained compressed data is composed of the smoothed signal, the corrected signal and the number of expansions, and the expanded data expanded using these is within the allowable error set from the original signal. On the other hand, in the range where the difference between the original signal and the value obtained by inversely transforming the smoothed signal and the detailed signal in the converted data does not exceed the desired tolerance, the extracted smoothed signal is obtained by thinning the data from the smoothed signal and the detailed signal The signal and the extraction detail signal are generated, and data including the number of expansions, the extraction smoothing signal, and the extraction detail signal is generated as compressed data obtained by compressing the original signal. Therefore, the obtained compressed data is composed of the extracted smoothed signal, the extracted detailed signal, and the number of expansions, and the expanded data expanded using these is within the allowable error set from the original signal. .

Further, the data converter of the present invention includes a wavelet converter for wavelet-converting an original signal with a desired number of expansions to generate converted data, a value obtained by inverse wavelet-converting the converted data, and an original signal. And a data correction unit that corrects the converted data so that the difference between the two becomes within a desired allowable error range and generates the original signal as compressed data. Therefore, the decompressed data obtained by decompressing the obtained compressed data is within the allowable error set from the original signal. Further, the data correction unit restores the original signal instead of the smoothed signal when the difference between the original signal and the value obtained by inverse wavelet-transforming the smoothed signal in the converted data exceeds the desired tolerance. The modified signal is generated, and the data including the modified signal, the smoothed signal, and the number of expansions is generated as compressed data. In this case, the obtained compressed data is composed of the smoothed signal, the corrected signal and the number of expansions, and the expanded data expanded using these is within the allowable error set from the original signal. Further, the data correction unit uses the smoothed signal and the detailed signal, respectively, within a range in which the difference between the original signal and the value inversely transformed from the smoothed signal and the detailed signal in the converted data does not exceed the desired tolerance. The extracted smoothed signal and the extracted detailed signal are generated by thinning out, and the data including the extracted smoothed signal and the extracted detailed signal and the number of times of expansion are generated as compressed data. In this case, the obtained compressed data consists of the extracted smoothed signal, the extracted detailed signal, and the number of expansions, and the expanded data expanded using these is within the allowable error set from the original signal. .

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. Embodiment 1 FIG. FIG. 1 is a configuration diagram showing a configuration when a data conversion device according to an embodiment of the present invention is used as a data compression device. In the figure, reference numeral 1 denotes an original signal input section, which is an original signal V composed of N pieces of data to be compressed.
₀ : Input as V ₀ (n) {n = 1, 2, ..., N}. Reference numeral 2 denotes a permissible error input unit that inputs the permissible range of the error between the original signal and the compressed / decompressed signal as the permissible error e. Reference numeral 3 denotes an expansion number determination unit, which inputs the expansion number J of the wavelet transformation. A wavelet transform unit 4 obtains a smoothed signal V _J using the original signal V ₀ and the number of expansions J by the following Expression 3. In the equation, k is the number of filter terms.

[0020]

(Equation 3)

Reference numeral 5 denotes a data correction unit, which uses the smoothed signal V _J , the original signal V ₀ , the allowable error e, and the expansion number J,
Modified signal VX ₀ : VX ₀ (n) {n consisting of N pieces of data
= 1, 2, ..., N}, the smoothed signal V _J, and the expansion count J
Is output. The data correction unit 5 includes a wavelet inverse transformation unit 5a and a correction signal generation unit 5b. Here, the wavelet inverse transform unit 5a uses the smoothed signal V _J and the number of expansions _J, and the inverse transform signal VI ₀ : VI ₀ (n) {n =
1, 2, ..., N} are output.

[0022]

(Equation 4)

The correction signal generator 5b also performs the following processing from n = 1 to n = N to generate and output the correction signal VX ₀ . First, the original signal V ₀ and the inverse conversion signal VI
₀ and the allowable error e are used to compare the n-th data of each of the original signal and the inversely converted signal. If the absolute value of the error does not exceed the allowable error e, the nth data of the correction signal is set to 0 (| VI ₀ (n) −V ₀ (n) | ≦ e
→ VX ₀ (n) = 0). When the absolute value of the error exceeds the allowable error e, the nth data of the corrected signal is set to the nth data of the original signal (| VI ₀ (n) −V ₀ (n) |
> E → VX ₀ (n) = V ₀ (n)). As a result of the above-described processing, the correction signal generation unit 5b creates and outputs a correction signal VX ₀ : VX ₀ (n) {n = 1, 2, ..., N} composed of N pieces of data.

As described above, according to the first embodiment, the allowable error can be set as follows. First, a portion where the difference between the inverse conversion signal VI ₀ and the original signal V ₀ does not fall within the desired tolerance e is a corrected signal V
Set the original signal to X ₀ . Then, the N correction signals VX ₀ , the smoothed signal V _J, and the number of expansions J thus obtained are used as compressed data. Therefore, the first embodiment
, No detail signal is required.

Therefore, the correction signals VX ₀ : VX ₀ (n) which are the compression signals obtained as a result from the data correction unit 5
The {n = 1, 2, ..., N} and the smoothed signal do not differ from the original signal V _{0 by} more than the allowable error e when they are put back together. That is, according to the first embodiment, the allowable error can be set when compressing / expanding data. In the first embodiment, the detailed signal is not taken into consideration as described above. When the signal to be compressed has a large proportion of low frequency components, the detailed signal is often not required to be compressed as a result. For such a signal, if the detailed signal is not used as described above, the calculation time is shortened and the compression rate is not lowered so much.

Embodiment 2 FIG. FIG. 2 is a block diagram showing the configuration of the data compression apparatus according to the second embodiment of the present invention. In the figure, reference numeral 24 is a wavelet transformation unit, which generates a smoothed signal V _J and a plurality of detailed signals W _{1 to} W _J by the following _equation 5 using the original signal V ₀ and the number of expansions J. In the equation, k is the number of filter terms.

[0027]

(Equation 5)

Reference numeral 25 is a data correction section, which uses the smoothed signal V _J , the plurality of detailed signals W _{1 to} W _J , the original signal V ₀ , the allowable error e, and the number of expansions J, and the smoothed signal and the plurality of signals. Extraction smoothing signal VX _{J is} obtained by extracting and leaving effective data for performing compression satisfying the allowable error in the detailed signal.
And output as a plurality of extraction detail signals WX _{1 to} WX _J.
Others are the same as FIG. Then, this data correction unit 25
Is a valid data updating unit 25a, as shown in FIG.
The effective data storage unit 25b, the wavelet inverse conversion unit 25c, the allowable error determination unit 25d, and the output unit 25e.

First, when the valid data updating unit 25a receives the control signal from the allowable error determining unit 25d, the extraction smoothing signal VX _J and the plurality of extraction detail signals WX _{1 to} WX _J are sent to the wavelet inverse conversion unit 25c. VX _J and WX ₁ to WX _J are created so that all the errors between the inversely converted signal VI ₀ obtained when input and the original signal V ₀ are within the allowable error range. Here, the valid data updating unit 25a sets the initial values to VX _J → V _J , WX _{1 to} WX _J → 0, and the inverse conversion signal is within the allowable error range for all the data of W _{1 to} W _J. Prioritize in order of importance. Then, each time it operates once, the data in W _{1 to} W _J are sequentially transferred to WX ₁ to WX _J one by one. The method of prioritizing is as follows.

First, let w _jn (m) be the weight of W _j (n) with respect to V ₀ (m). For example, when j = 2, the weight is derived as shown in the following equation 6 by utilizing the equation 8 described later.

[0031]

(Equation 6)

Next, x _jn shown in the following _Expression 7 is defined as an evaluation amount for prioritizing W _j (n).

[0033]

(Equation 7)

Then, x _jm is calculated over j = 1 to J and l = 1 to N, and a priority is given from a coefficient having a larger value. Note that x _jn = | W _j (n) | is defined and x _jn is j =
1 to J, n = 1 to N may be calculated, and the priority may be given from a coefficient having a larger value.

Next, the effective data storage unit 25b, when stores the VX _J and WX ₁ ~WX _J obtained from valid data updating unit 25a, the next effective data updating unit 25a is operated, these Is transferred to the valid data updating unit 25a. Next, the wavelet inverse transformation unit 25c is composed of N pieces of data using the following equation 8 according to the extraction smoothing signal VX _J , the plurality of extraction detail signals WX _{1 to} WX _J , and the number of expansions J. The inverse conversion signal VI ₀ : VI ₀ (n) is created.

[0036]

(Equation 8)

Then, the allowable error judgment unit 25d outputs the original signal V ₀ , the allowable error e, and the converted signal VI ₀ when all the errors between the original signal and the inverse converted signal are within the allowable range. A control signal for operating the section 25e is output. On the other hand, if it is not within the allowable error range, a control signal for operating the valid data updating unit 25a is output. The output unit 25e then includes the allowable error determination unit 25d.
When there is a command from the output, the extraction smoothing signal VX _J and a plurality of extraction detail signals WX _{1 to} WX _J are output.

As described above, in the second embodiment, the detailed signals (extracted detailed signals WX _{1 to} WX _J ) obtained by the wavelet transform are also used as compressed data. That is, in the second embodiment, the allowable error determining unit 25d allows all the inverse conversion signals VI ₀ obtained by inverse conversion using the extraction smoothed signal VX _J and the plurality of extraction detail signals WX _{1 to} WX _J. The extracted smoothed signal VX _J and the plurality of extracted detailed signals WX _{1 to} WX _J are selected so as to be within the error e,
These are used as compressed data. As a result, the data can be compressed within the set allowable error range, and the compression rate can be further increased. In the second embodiment, the coefficients are prioritized, but this is necessary for extracting the coefficients required for compression at a higher speed.

Embodiment 3 3 (a) and 3 (b) are configuration diagrams showing the configuration of the data conversion apparatus according to the third embodiment of the present invention. In the figure, 31a, 31b
Is a compressed data input section, 32 is a wavelet inverse transformation section,
Reference numeral 33a is a decompression signal generation unit, and 33b is a wavelet inverse transformation / decompression signal generation unit. FIG. 3A corresponds to the above-described first embodiment, and expands the compressed data obtained in the first embodiment. Further, FIG. 3B corresponds to the above-described second embodiment and expands the compressed data obtained in the third embodiment.

First, the operation of the data converter shown in FIG. 3A will be described. The compressed data input unit 31a inputs the correction signal VX ₀ , the smoothing signal V _J , and the expansion count J. Further, the wavelet inverse transform section 32 uses the smoothed signal V _J and the expansion count J to create the inverse transform signal VI ₀ : VI ₀ (n) consisting of N pieces of data by using the above _{equation 8.} . In this case, the smoothed signal V _J is regarded as the extracted smoothed signal VX _J in which all the elements are extracted.

Then, the expansion signal generator 33a uses the modification signal VX ₀ and the inverse conversion signal VI ₀ to modify the modification signal VX _0.
If the n-th data of 0 is 0, the expanded signal VR _{0 to be} output
The n-th data of is the n-th data of the inverse conversion signal (VX ₀ (n) = 0 → VR ₀ (n) = VI ₀ (n)). Further, expansion signal generating unit 33a, when the n-th data correction signal VX ₀ is not 0, n of the extension signal VR ₀ to be output
The nth data is the nth data of the correction signal VX ₀ (VX ₀ (n) ≠ 0 → VR ₀ (n) = VX ₀ (n)). Then, the decompression signal generation unit 33a outputs these from n = 1 to n.
= Expansion signal VR consisting of N data
₀ : VR ₀ (n) {n = 1, 2, ..., N} is created.

Next, the operation of the data converter shown in FIG. 3B will be described. The compressed data input unit 31b inputs the extracted smoothed signal VX _J , the plurality of detailed signals WX _{1 to} WX _J , and the expansion count J. Further, the wavelet inverse transform / expansion signal generation unit 33b outputs the input extracted smoothed signal VX _J , the plurality of detailed signals WX _{1 to} WX _J , and the expansion count J,
By performing the inverse wavelet transform using the above equation 8, N
Expanded signal VR ₀ (n): VR ₀ (n) composed of individual pieces of data
Create {n = 1, 2, ..., N}. As described above, according to the third embodiment, the first and second embodiments
The signal compressed at can be decompressed.

As the numerical value (k = 8) of the filter in the above formula, those shown in Table 1 below can be cited as an example.

Fourth Embodiment FIG. 4 is a configuration diagram showing a configuration of a data compression device according to the fourth embodiment of the present invention. In the figure, 41 is an original signal input section,
Original signal V ₀ : V ₀ (n) {n = 1,
The compressed signal of 2, ..., N} is input. A permissible error input unit 42 inputs the permissible error range between the original signal V ₀ and the compressed / expanded signal as the permissible error e. Reference numeral 43 denotes an expansion number distribution unit, which determines and outputs a combination of the expansion numbers of the wavelet transform. Reference numeral 44 is a wavelet transform unit, which uses the original signal V ₀ input by the original signal input unit 41 and the expansion count J determined by the expansion count distribution unit 43 to generate a smoothed signal V _J by the following Expression 9. create.

[0045]

(Equation 9)

Reference numeral 45 is a data correction unit, which uses the smoothed signal V _J , the original signal V ₀ , the allowable error e, and the number of expansions J to make a correction signal VX ₀ : V composed of N pieces of data.
X ₀ (n) {n = 1, 2, ..., N} is output. The wavelet converter 44 and the data corrector 45 constitute a data converter 40. The data converter 40
Are arranged in plural. Reference numeral 46 is a compressed data number comparison unit. The compressed data number comparison unit 46 calculates and compares the sum of the data numbers of the smoothed signal V _J and the modified signal VX ₀ output from each data conversion unit 40 for each expansion number J, and compares the sum of the data numbers to obtain the most data number. The number of expansions J when the number is small, and the smoothing signal V _J and the correction signal VX ₀ at that time are output.

Further, as shown in FIG. 4B, the data correction section 45 is composed of a wavelet inverse conversion section 45a and a correction signal generation section 45b. The wavelet inverse transformation unit 45a uses the smoothed signal V _J and the expansion count J,
The following _equation 10 outputs the inverse conversion signal VI ₀ : VI ₀ (n) {n = 1, 2, ..., N} composed of N pieces of data.

[0048]

(Equation 10)

On the other hand, in the modified signal generator 45b, first,
It is determined whether or not the absolute value of the error between the n-th data of the original signal V ₀ and the inverse conversion signal VI ₀ is within the allowable error e. In this determination, if the absolute value of the error does not exceed the allowable error e, the correction signal VX _{0 to be} output.
The n-th data of is set to 0 (| VI ₀ (n) -V
₀ (n) | ≦ e → VX ₀ (n) = 0). On the other hand, when the absolute value of the error exceeds the allowable error e, the nth data of the modified signal VX _{0 is set as} the nth data of the original signal V ₀ (| VI ₀ (n) −V ₀ (n) │ <e → VX ₀ (n) = V
₀ (n)). Then, the above processing is performed from n = 1 to n = N, and the correction signals VX ₀ : VX composed of N pieces of data are obtained.
₀ (n) {n = 1, 2, ..., N} is output.

That is, as described above, in the fourth embodiment, the wavelet converter 4 shown in FIG. 1 is used.
And a plurality of data correction units 5 are provided. Then, by setting a predetermined number of decompressions in each wavelet conversion unit and comparing the obtained number of compressed data, the number of decompression with the smallest number of compressed data and the compressed data at that time are obtained. Is.

In the first and second embodiments, it is necessary to appropriately determine the number of wavelet expansions.
However, if the number of expansions is large, a good compression result may not always be obtained. The number of decompression that maximizes the compression rate is not known in advance, so if the number of decompression is not optimal, the compression rate may not be sufficiently high.
In the fourth embodiment, an appropriate expansion range or combination is determined in advance to automatically determine the expansion frequency with the highest compression rate and compress the data. Therefore, the work of strictly determining the number of wavelet expansions can be omitted, and the work efficiency can be improved. Also, the calculation time can be shortened.

Embodiment 5 FIG. By the way, although the plurality of wavelet conversion units 4 and the data correction unit 5 in FIG. 1 are provided in the fourth embodiment, the wavelet conversion unit 24 and the data correction unit 25 in FIG.
You may make it equipped with two or more. By doing this,
The compression rate can be made higher than that in the fourth embodiment.

Embodiment 6 FIG. FIG. 5 is a configuration diagram showing the configuration of the data compression apparatus according to the sixth embodiment of the present invention. In the figure, the original signal input unit 1, the allowable error input unit 2, the wavelet conversion unit 4, and the data correction unit 5 are the same as those in FIG. In FIG. 5, reference numeral 53 is a maximum decompression number setting unit for inputting the maximum value J of the decompression number of the desired wavelet conversion at the time of data compression.

Reference numeral 55 denotes a compressed data storage unit, which uses the decompression number d, the smoothed signal V _d , and the correction signal VX _{0 (d)} as the compressed data at the decompression number d, and the compressed data at various decompression numbers d. Remember. Further, reference numeral 56 is a development counter, which increments d by 1 each time the sequential processing is performed once with an initial value of d = 1 and outputs.

Reference numeral 57 denotes an expansion number determination unit, which uses the expansion number d from the expansion number counter 56 and the maximum value J specified from the maximum expansion number setting unit 53. First, when d is smaller than J, this expansion is performed. The number of times d is sent to the wavelet converter 4. On the other hand, when d becomes equal to J, the control signal s of the operation start command is output to the compressed data number comparison unit 59.
The compressed data number comparison unit 59 that has received the operation start command receives the smoothed signal V _d stored in the compressed data storage unit 55,
Using the correction signal VX _{0 (d)} and the number of expansions d, V _d and VX
The sum of the data numbers of _{0 (d)} is calculated for each d and compared. Then, the compressed data number comparison unit 59 outputs the number of expansions d when the number of data is the smallest, and the smoothing signal Vd and the correction signal VX _{0 (d)} at that time.

Also in the sixth embodiment, similarly to the fourth embodiment, by determining an appropriate range or combination of the number of expansions, the number of expansions with the highest compression rate is automatically determined. And compress the data.
Therefore, the work of strictly determining the number of wavelet expansions can be omitted, and the work efficiency can be improved. In addition, according to the sixth embodiment, the fourth and fifth embodiments
The device configuration can be made smaller than that of.

Embodiment 7 FIG. By the way, in the sixth embodiment, in obtaining the effective data (correction signal) by performing the wavelet conversion, the first embodiment shown in FIG.
However, it goes without saying that this may be the same as in the second embodiment shown in FIG. By doing so, according to the seventh embodiment, the detailed signal by the optimum expansion order can be taken into consideration, so that the compression rate can be made higher than in the sixth embodiment.

Embodiment 8. Further, in obtaining the effective data, in the second, fifth, and seventh embodiments, the effective data extracted from the entire conversion result is used based on the entire conversion result and the original signal. You may make it extract more effective data than a part. FIG. 6 is a block diagram showing the structure of the data compression apparatus according to the eighth embodiment. In the figure, the wavelet conversion unit 4 is the same as that of FIG. 2, and the original signal input unit 1, the allowable error input unit 2, the maximum expansion number setting unit 53, the expansion number counter 56, the expansion number determination unit 57, and The compressed data number comparison unit 59 is the same as that shown in FIG.

In FIG. 6, reference numeral 61 is a data correction unit. In the data correction unit 61, first, the smoothed signal V
_d , the detailed signal W _d , the original signal V ₀ , the allowable error e, the expansion number d, and the extracted detailed signals WX _{1 to} WX _d-1 already stored in the compressed data storage unit 62 are input. Then, the data correction unit 61 determines the smoothed signal V _d and the detailed signal W.
Of _d , effective data for compression that satisfies the allowable error e is extracted, and the remaining data is output as the extracted smoothed signal VX _d and the extracted detailed signal WX _d , respectively. on the other hand,
The compressed data storage unit 62 stores the remaining extracted smoothed signal VX _d , extracted detailed signal WX _d, and the number of expansions d as one set.

The data correction section 61 will be described in more detail below. As shown in FIG. 6B, the data correction unit 61 includes a valid data updating unit 61a and a valid data storage unit 61.
b, wavelet inverse transformation unit 61c, allowable error determination unit 6
1d and an output unit 61e. First, the valid data updating unit 61a adjusts all errors between the inverse conversion signal VI ₀ and the original signal V ₀ within an allowable range.
VX _d and WX _d are generated. This inverse conversion signal VI
₀ is obtained when the extracted smoothed signal VX _d and the plurality of extracted detailed signals WX _{1 to} WX _d output by the valid data updating unit 61 a are input to the wavelet inverse transformation unit 61 _c .

In the valid data updating unit 61a, first, the initial values are set to VX _d → V _d and WX _d → 0 so that the inverse conversion signal VI ₀ falls within the allowable range for all the data in W _d. Prioritize in order of importance. Then, each time the operation is performed, the data in W _d are sequentially transferred to WX _d one by one. Here, the priority order is, as described above, the valid data updating unit 2 shown in FIG.
Attach it in the same way as 5a.

The valid data storage unit 61b stores the output from the valid data updating unit 61a, and when the valid data updating unit 61a operates next time, the stored data is stored in the valid data updating unit 61a. 61a. next,
The wavelet inverse transform unit 61c is configured to smooth the smoothed signals VX _d ,
Inverse conversion signal VI ₀ : VI ₀ (n) {n = 1, 2, ..., Consisting of N pieces of data by using a plurality of detailed signals WX _{1 to} WX _d and the number of expansions, using the following equation 11. N} is created.

[0063]

[Equation 11]

The permissible error judgment unit 61d judges the error between the original signal V ₀ and the inverse conversion signal VI _0, and when this is not within the range of the permissible error e, the valid data updating unit 61a.
And outputs a control signal for operating the. By the above,
Until all the errors between the original signal V ₀ and the inverse conversion signal VI ₀ fall within the allowable error e, the valid data updating unit 61a sequentially moves the data in W _d one by one to WX _d . Then, the permissible error determination unit 61d outputs a control signal for operating the output unit 61e when the errors between the original signal V ₀ and the inverse conversion signal VI ₀ are all within the permissible error e. As a result, the output unit 61e outputs the extracted smoothed signal VX _d and the plurality of extracted detailed signals WX _{1 to} WX _d .

Using this output signal, the compressed data number comparison unit 59 uses the extracted smoothed signal VX _d and the plurality of extracted detailed signals stored in the compressed data storage unit 62, as in the seventh embodiment. using the WX ₁ ~WX _d, VX _d and WX
The sum of the data numbers _{1 to} WX _d is calculated for each d and compared. Then, the compressed data number comparison unit 59 outputs the number of expansions when the number of data is the smallest, and the smoothing signal and the correction signal at that time. That is, according to the eighth embodiment, when the number of expansions is J, the extracted smoothed signal and the extracted detailed signal are obtained from the smoothed signal and the detailed signal, and the extracted detailed signal obtained up to the number of expansions J−1 is obtained. The added data is used as compressed data when the expansion count is J. For this reason, in the eighth embodiment, since the valid data is successively selected for each number of expansions, the calculation speed is higher than that in the seventh embodiment.

Embodiment 9 FIG. 7 is a configuration diagram showing the configuration of a data compression device according to the ninth embodiment of the present invention. In the figure, the original signal input unit 1, the allowable error input unit 2, the wavelet conversion unit 4, and the data correction unit 5 are the same as those in FIG.

In FIG. 7, reference numeral 71 denotes a compressed data updating unit, which is a correction signal VX _{0 (d)} output by the data correction unit 5 when the number of expansions is _d and a smoothed signal V output by the wavelet conversion unit 4. _It stores _d and the latest compressed data and outputs Vd _-1 and VX _{0 (d-1)} which are the previous compressed data.
Reference numeral 72 is a development counter, which has an initial value of d = 1.
This is output as the data number determination unit 7 described below.
Each time 5 operates once, the value of d is incremented by 1 and output.

Reference numeral 73 denotes a data number counter, which smoothes the signal V _d and corrects the signal VX _{0 (d)} at the expansion number d.
Then, the sum of the data numbers of is calculated, and this is output as the compressed data number DL _d . A data number storage unit 74 obtains the current compressed data number DL _d from the data number counter 74 at the decompression number _d and stores this, and instead stores the previous compressed data DL _d-1 . Output.

[0069] Then, the data count judgment unit 75 compares the compressed data DL _d-1 of this compressed data count DL _d the previous in development times d, if DL _d is larger than DL _d-1, is updated The expanded number of times d + 1 is sent to the wavelet conversion unit 4, and in other cases, the control signal s is output to the compressed data output unit 76 for operation. When the compressed data output unit 76 receives the operation command from the data determination unit 75, the previous compressed data output from the compressed data updating unit 71 and the previous expansion count output from the expansion count counter 72 are output. It

In the ninth embodiment, the number of expansions that maximizes the compression rate is automatically determined without compressing the number of expansions, and the data is compressed. That is, as described above, the compression is continued while increasing the number of expansions until the compression ratio becomes poor. Therefore, it is possible to obtain the compressed data with the optimum expansion order only by inputting the original signal and the allowable error without setting the maximum expansion number. Therefore, the work of strictly determining the number of wavelet expansions can be omitted, and the work efficiency can be improved.

Tenth Embodiment By the way, in the ninth embodiment, the wavelet conversion unit and the data correction unit may have the same configurations as those of the second embodiment shown in FIG. 2 or shown in FIG. The same as in the eighth embodiment. By doing so, the finally obtained compressed data takes into consideration the detailed signal by the optimum number of expansions (degrees), so the compression rate should be made higher than in the ninth embodiment. You can

[0072]

Embodiments of the present invention will be described below.
The compressed / decompressed data is used for various purposes such as data analysis, control, abnormality diagnosis / detection, and the like, but the case of abnormality detection will be described below as an example. Here, anomaly detection in a certain manufacturing process using a blast furnace is taken as an example.
This manufacturing process has a problem that when a phenomenon in which the temperature in the blast furnace rapidly changes appears, defective products are started to be produced. In order to reduce the rate of producing this defective product, it is necessary to quickly detect an abnormality in temperature change in the blast furnace.

FIG. 8 is a characteristic diagram showing 7200 data obtained by measuring the temperature in the blast furnace at 1-minute intervals from 00:00 on the 1st day of 1995 to 23:59 on the 5th day of the same month. In order to detect the temperature change abnormality in the blast furnace earlier, the characteristics are actually analyzed from past historical data. The past history data is composed of the temperature in the blast furnace and the process state (abnormal or normal) at the time of measurement.

That is, a large amount of data as shown in FIG. 8 is held, and this feature is analyzed as shown below.
When "1 minute temperature change" is defined as the temperature speed, the difference between the average temperature speed from a certain time t to 10 minutes before this time and the average temperature speed from the time t to 10 minutes after this time. However, if it is 0.1 (° C./min) or more, the process is in an abnormal state. The data in FIG. 8 is measured at 1-minute intervals, and by first-order differentiating the data, data representing the temperature velocity is obtained as shown in FIG. 9 (a). No abnormalities appeared in the process during this measurement period.

In this first-order differentiated data, before and after 1
By taking the difference of 0 averages, the data shown in FIG. 9B is obtained, and this represents the difference of the averages of the temperature velocity before and after 10 minutes. As is clear from FIG. 9B, it is found that the conclusion is not inconsistent even when the above-mentioned analysis result is compared with this data.

However, in general, it is not preferable to hold a large amount of data as shown in FIG. 8 described above in order to configure the device. In order to hold a large amount of data, it is necessary to prepare as many storage units as that, and there are many problems such as the cost of the device. For this reason, it is usual that the above-mentioned data shown in FIG. 8 is compressed and held, and is expanded and used at the time of analysis.

Here, the data shown in FIG. 8 is compressed by the conventional method with a permissible error of 3%, and is expanded, as shown in FIG. 10 (a). And this FIG.
If data corresponding to the average difference of 10 minutes before and after the temperature velocity is obtained based on the extension data of 0 (a), as described above,
It becomes as shown in FIG. In this,
There is a contradiction with the analysis result from the data in FIG. This is because the conventional compression / expansion includes a property that the original signal does not have, and a sudden change that does not exist in the original signal appears here.

On the other hand, as shown in FIG. 11A, the data compressed / decompressed with the allowable error of 3% by the data compression device according to the present invention is almost the same as the original data of FIG. To be done. This has an average error of 1.62% and a compression ratio of 154 times. And this FIG.
The difference between the averages of 10 data before and after the first derivative of the data shown in 1 (a) is shown in FIG. 11 (b), which is consistent with the conclusion based on the data shown in FIG. That is, the present invention is effective for compression / expansion of original data in abnormality diagnosis for observing signal discontinuity.

FIG. 12 is data showing the result of compressing / expanding data by the data compressing apparatus according to the present invention. FIG. 12 (a) shows the number of expansions of 2, the average error is 0.003% and the compression ratio is 0.003%. Was 1.99 times. In both cases, the allowable error was 3%. In FIG. 12B, the number of expansions was 9, the average error was 0.343%, and the compression rate was 147 times.

[0080]

As described above, according to the present invention, the wavelet-converted data is inversely wavelet-converted once, and it is determined that the original data is within the allowable error. . Therefore, it is possible to set the allowable error in converting the data, and for example, it is guaranteed that the error between the compressed and expanded data and the original data is within the allowable range, and higher reliability is obtained. It has the effect of being obtained.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a configuration when a data conversion device according to an embodiment of the present invention is used as a data compression device.

FIG. 2 is a configuration diagram showing a configuration of a data compression device according to a second embodiment of the present invention.

FIG. 3 is a configuration diagram showing a configuration of a data compression device according to a third embodiment of the present invention.

FIG. 4 is a configuration diagram showing a configuration of a data compression device according to a fourth embodiment of the present invention.

FIG. 5 is a configuration diagram showing a configuration of a data compression device according to a sixth embodiment of the present invention.

FIG. 6 is a configuration diagram showing a configuration of a data compression device according to an eighth embodiment of the present invention.

FIG. 7 is a configuration diagram showing a configuration of a data compression device according to a ninth embodiment of the present invention.

FIG. 8 is a characteristic diagram showing data obtained by measuring the temperature in the blast furnace at 1-minute intervals.

9 is a characteristic diagram obtained by first-order differentiating the data in FIG.

FIG. 10 is a characteristic diagram showing a result of compressing the data of FIG. 8 by a conventional method and expanding the data.

FIG. 11 is a characteristic diagram showing data compressed / decompressed according to the present invention.

FIG. 12 is a characteristic diagram showing data compressed / decompressed according to the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Original signal input part, 2 ... Permissible error input part, 3 ... Expansion number determination part, 4 ... Wavelet conversion part, 5 ... Data correction part, 5a ... Wavelet inverse conversion part, 5b ... Correction signal generation part.

Claims (16)

[Claims] 1. An original signal is wavelet-transformed a desired number of times to generate transformed data, and a difference between a value obtained by inverse wavelet-transforming the transformed data and the original signal is within a desired allowable error range. A method of converting data, characterized in that the converted data is modified to generate compressed data obtained by compressing the original signal. 2. The data conversion method according to claim 1, wherein when a difference between a value obtained by inverse wavelet-transforming the smoothed signal in the converted data and the original signal exceeds a desired tolerance, A data conversion characterized by generating a modified signal returned to the original signal instead of the smoothed signal, and generating data consisting of the number of expansions, the smoothed signal and the modified signal as compressed data obtained by compressing the original signal. Method. 3. The data conversion method according to claim 1, wherein the difference between the original signal and the value obtained by inversely converting the smoothed signal and the detailed signal in the converted data does not exceed a desired tolerance. The extracted smoothed signal and the extracted detailed signal are generated by thinning out the data from the smoothed signal and the detailed signal, respectively, and the data consisting of the number of expansions, the extracted smoothed signal and the extracted detailed signal is compressed by compressing the original signal. A data conversion method characterized by generating as data. 4. The data conversion method according to claim 2 or 3, wherein the compressed data is generated with a plurality of decompression times, and compression is performed when the number of decompression times is the smallest among the obtained plurality of compressed data. A data conversion method characterized by selecting data. 5. The data conversion method according to claim 2 or 3, wherein the compressed data is simultaneously generated with a plurality of decompression times, and the decompression number with a small number of compressed data among the obtained plurality of compressed data is used. A data conversion method characterized by selecting compressed data. 6. The data conversion method according to claim 2, wherein compressed data is generated by increasing the number of expansions by a predetermined number,
If the number of compressed data obtained by the expansion number set this time is greater than the compressed data obtained by the previously set expansion number, select the compressed data at the previously set expansion number. Characteristic data conversion method. 7. The data conversion method according to claim 2, wherein the smoothed signal and the number of times of expansion are used to inversely transform the wavelet to generate an inverse transformed signal, and the inverse transformed signal and the modified signal are expanded. A data conversion method characterized by generating a signal. 8. The data conversion method according to claim 2, wherein an expanded signal is generated by performing inverse wavelet conversion using the extracted smoothed signal and the extracted detail signal. 9. A wavelet transform section for wavelet transforming an original signal with a desired number of expansions to generate transformed data, and a difference between a value obtained by inverse wavelet transforming the transformed data and the original signal is a desired tolerance. A data conversion device, comprising: a data correction unit that corrects conversion data so as to be within an error range and generates original data as compressed data that is compressed. 10. The data conversion apparatus according to claim 9, wherein the data correction unit determines a desired tolerance by a difference between a value obtained by inverse wavelet-transforming a smoothed signal in the conversion data and the original signal. If it exceeds, a correction signal returned to the original signal is generated instead of the smoothed signal, and data including the correction signal, the smoothed signal, and the number of expansions is generated as compressed data. Converter. 11. The data conversion apparatus according to claim 9, wherein the data correction unit has a desired tolerance for a difference between a value obtained by inversely converting a smoothed signal and a detailed signal in the converted data and the original signal. Within the range not exceeding, the extracted smoothed signal and the extracted detailed signal are generated by thinning out the data from the smoothed signal and the detailed signal, respectively, and the data including the extracted smoothed signal and the extracted detailed signal and the number of times of expansion are generated. A data conversion device characterized by being generated as compressed data. 12. The data conversion device according to claim 10, further comprising a plurality of sets of the wavelet conversion section and the data correction section, and determining a combination of wavelet conversion expansion times used in each of the wavelet conversion sections. A data conversion device, comprising: a decompression number distribution unit, and a compressed data number comparison unit that selects and outputs the compressed data having the smallest number of compressed data from the data correction unit. . 13. The data conversion apparatus according to claim 10, wherein a maximum expansion number setting unit that sets a plurality of expansion times to the wavelet conversion unit within a predetermined maximum expansion number range, and the plurality of expansions. A data conversion device, comprising: a determination unit that selects and outputs the compressed data having the smallest number of compressed data among the compressed data obtained each time. 14. The data conversion device according to claim 10, wherein the expansion number counter that sets a plurality of expansion numbers, which are increased by a predetermined number, to the wavelet conversion unit and the expansion number set this time are obtained. Compressed data is
When the number of compressed data obtained by the previously set decompression number is larger than that of the compressed data, the data conversion device is provided with a determination unit for selecting and outputting the compressed data having the previously set decompression number. . 15. The data conversion apparatus according to claim 10, wherein the wavelet inverse conversion unit generates an inverse conversion signal by performing an inverse wavelet conversion using the smoothed signal and the number of expansions, and the inverse conversion signal. And a decompression signal generation unit for generating and outputting a decompression signal by the correction signal and the correction signal. 16. The data conversion apparatus according to claim 11, further comprising a decompression signal generation unit for generating and outputting a decompression signal by performing inverse wavelet conversion using the extracted smoothed signal and the extracted detail signal. A data conversion device characterized by the above.