JPH11195991A - Method and device for compressing and decompressing analog signal data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To conduct compression/decompression of analog signal data effectively by predicting a next sampling value from a certain sample value and a sample just preceding the sample value and calculating an error between the predicted value and the actual sample. SOLUTION: Obtained analog data are expressed as a power series expansion equation using sampling times t0 -t3 for variables, and a predicted value of a sample to be taken next is calculated by using the equation. Since an error between the predicted value and the sample obtained actually is very small, the data are compressed very effectively. Other wise quadratic power series expansion equation using the sampling times t0 -t3 for variables is used to calculate a predicted value of the sample. Furthermore, relative values S1 -S3 are obtained, where the 1st sample is set to 0.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a technique for compressing data useful for transmitting and recording various analog data and a technique for restoring the compressed data.

[0002]

2. Description of the Related Art An analog telemeter for analog data such as voltage, current and power, or data such as temperature and humidity, and various weather data, and an analog such as telecontrol for analog values such as temperature, pressure, flow rate, rotation speed, and position. When transmitting data, recording these data, or restoring the recorded data to the original analog value, generally, the original analog value is sampled, converted to a digital code, and this data is converted to normal data. Depending on the transmission method,
Or to record. In such a case, error control is usually performed to correct a code error occurring in a transmission medium or a recording medium, but in order to perform this error control, a redundant code for error control is added to a data code. Therefore, reductions in transmission efficiency and recording efficiency due to this cannot be avoided. Furthermore, if one system attempts to handle data of many channels, the signal speed of the data must be increased, and the higher the speed, the greater the number of code errors. In order to prevent this, it is necessary to take measures such as limiting the signal speed by reducing the amount of data to be handled or increasing the signal power, but increasing the signal power is generally difficult in many cases.

[0003] Therefore, in general, there is a large degree of redundancy between adjacent sample values which is peculiar to analog data. Therefore, it is conceivable to compress the data by removing this redundancy. As a method for this, first, by calculating the difference with respect to the immediately preceding sample value,
There is a method of reducing the average level for expressing data, compressing the data, and accumulating the difference value during reproduction to restore the original data value.

[0004] Although a considerable compression effect can be obtained by this method, there is the following method as a method of obtaining a compression effect even more than this method. It uses two sample values, the sample value of interest and the sample value immediately before it, to predict the value that the next sample will take from an extension line connecting the two sample values with a straight line, This is a method of compressing data by calculating the error between the sample value and the actual sample value for successive sample values, and the compression effect is greater than when the difference between the former and the previous sample value is used. is there.

[0005]

An effective compression can be obtained by the above-mentioned method using two sample values of the sample value of interest and the sample value immediately before it, because the next sample value is the previous two sample values. Obviously, this is the case when the value is on an extension of a straight line connecting the values. However, in actual data, it is rare that the data is on such an extension line, and in general, there are many bends and quadratic or exponential data, so that the compression effect is still not sufficient. It was hard to say.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to find a method and apparatus for more effectively compressing and restoring data of an analog signal.

[0007]

According to a first aspect of the present invention, there is provided a method for compressing and restoring an analog signal, comprising sampling and sampling a continuously changing analog signal at a predetermined timing, compressing the sampled signal, and compressing the sampled signal. A method of decompression, on the compression side, using a power series expansion formula with the sampling timing as a variable, calculate a predicted value of a sample value that will be taken next based on a plurality of continuous sample values,
Calculate the error between the predicted value and the sample value actually obtained,
On the reconstruction side, a predicted value of a sample value to be taken next is calculated based on a plurality of sample values sequentially calculated, using a power series expansion formula with sampling timing as a variable, and the predicted value and It is characterized by restoring the original sample value from the error.

[0008] The compression of an analog signal according to the invention of claim 2
・ The restoration method is to convert a continuously changing analog signal to a predetermined
Sampling and sampling at the timing, compress this
After the compression, the compression side
Second-order power series expansion formula with the pulling timing as a variable
, Two consecutive sample values s_n-2, S_n-1Based on
Next sample value s_nPredicted value p_nTo the following equation (1)
And the predicted value p_nAnd the sample actually obtained
Value s_nError x_nIs calculated by the following equation (2) and restored.
Side, the sampling timing is used as a variable.
Using the following power series expansion formula, two markers calculated sequentially
This value s_n-2, S_n-1The next sample value s based on_n
Predicted value p_nIs calculated by equation (1), and the
Predicted value p _nAnd prediction error x_nFrom the original sample value s_nRestore
It has a characteristic in its origin.

_Pn = 3 (s _n−1 −s _n−2 ) (1)

[0010] _{x n = s n -3 (s} n-1 -s n-2) ... (2)

S _n = x _n +3 (s _n−1 −s _n−2 ) (3)

According to a third aspect of the present invention, there is provided the analog signal data compression / decompression method according to the first or second aspect, wherein the sample value is obtained as a relative value with the first value being 0. Has features. Hereinafter, the present invention will be described more specifically.

As is well known, many curves y can be represented by the following power series.

Y = a ₀ + a ₁ x + a ₂ x ² +... _An x ⁿ +.

A ₀ , a ₁ , a ₂ ,...,
Symbols such as a _n ,... are coefficients for each order that specifies the curve y. In most cases, the higher the order of these coefficients, the smaller their values.

The original analog data in the present invention is also represented by a continuous curve. Therefore, here, first, a method of expanding the obtained signal into a power series up to the second order using the sampling time as a variable will be specifically described (the invention of claim 2).

When developing into a power series up to the second order, three consecutive sample data are used. In this case, as shown in FIG. 1, at the sampling timing from t ₀ to t ₁ , t ₂ ,..., The sample value data is set to relative value data s ₀ , s ₁ , s ₂ , s ₀ = 0. It is desirable to convert to ... (the invention of claim 3). The time is considered as a relative time based on t ₀ , and the calculation is performed on this relative value as described later. That is, as shown in FIG. 1, the calculation at this time is performed on the value when s ₀ is the origin of the coordinates. And data up to s ₂ are already to be transmitted Zumi, s _0, s _1, s _2, s ₃ using three relative value data obtains a prediction error for. Hereinafter, the procedure will be described using a case where it is used for data transmission as an example.

The time scale is t ₀ = 0, t ₁ = 1, t ₂
Are represented by integers such as 2... S ₀ , s ₁ ,
s ₂ is represented by the sum of a linear expression and a quadratic expression, and there is no third-order or higher component. If a is a coefficient of a linear equation and b is a coefficient of a quadratic equation,
s ₁ and s ₂ are respectively expressed as follows.

S ₁ = 1 ¹ a + 1 ² b = a + b

S ₂ = 2 ¹ a + 2 ² b = 2a + 4b

Then, using this as a simultaneous equation, the coefficient a,
When b is obtained, it becomes as follows.

[0022] _{a = 2s 1 - (s 2} /2)

[0023] _{b = (s 2/2)} -s 1

Once these two coefficients are obtained, they can be used to determine the predicted value p ₃ that the relative value data s ₃ will take at the next sampling time t ₃ , as shown in the following equation: Become.

P ₃ = 3 ¹ a + 3 ² b = 3a + 9b = 3 (s ₂ −s ₁ ) Therefore, the prediction error x _{3 for} the actual relative value data s ₃ can be obtained as follows: x ₃ is transmitted to the receiving side.

X ₃ = s ₃ -p ₃ = s ₃ -3 (s ₂ -s ₁ )

On the receiving side, the transmitted prediction error x ₃
Using the relative value data s ₁ and s ₂ already restored, the following operation is performed to restore s ₃ .

S_Three= X_Three+ P_Three= X_Three+3 (s_Two−s₁Actually, as described later, sample value data is restored
And the relative value data s₁, S_Two
, So s₀Original sample data and predicted values of
P_ThreeAnd the transmitted error x_ThreeOf the three
And s _ThreeIs restored. I
Therefore, the subject of the operation on the transmitting side is x_Three= S_Three-P_ThreeIn
And the subject of the operation on the receiving side is s_Three= X_Three+ P_ThreeIt is.

As described above, when the prediction error of s ₃ is calculated, s ₀ is set as the origin. Similarly, when the prediction error of s ₄ is calculated, s ₁ may be set as the origin. The calculation of the predicted value for successive sample values is performed while shifting the origin one sampling time at a time.

As described above, according to the first aspect of the present invention, on the compression side, a predicted value of a sample value to be taken next is calculated by using a power series expansion equation using a sampling time as a variable, and this predicted value is calculated. Since the error between the value and the actually obtained sample value is calculated, the prediction accuracy is higher and the error is reduced as compared with a method of calculating the error by taking a predicted value on an extension of two sample values. Therefore, there is an effect that the data amount is significantly reduced. Therefore, when transmitting or recording such data, a large amount of data can be quickly handled because the data amount is small.

According to the second aspect of the present invention, on the compression side, the predicted value of the sample value to be taken next is calculated using the quadratic power series expansion equation using the sampling time as a variable. In addition to the effects of the first aspect, there is an effect that the calculation method and the arithmetic circuit configuration thereof can be simplified.

Further, according to the third aspect of the present invention, since the prediction error is calculated by a relative value with the first sample value being 0, the prediction error is calculated in the compression device or the decompression device rather than using the sample value as it is. In this case, the amount of data to be handled is reduced, and more rapid calculations can be performed.

A method using a cubic power series expansion equation is as follows. In this case, s ₀ , s ₁ ,
using four successive data s _2, s ₃ obtains the prediction error for s _4. As before, in this case 4
Assuming that there are no components higher than the first order, assuming that the first-order, second-order, and third-order coefficients are a, b, and c, respectively, s ₁ , s ₂ , and s ₃ are expressed by the following ternary simultaneous equations, respectively. .

S ₁ = 1 ¹ a + 1 ² b + 1 ³ c = a + b + c

S ₂ = 2 ¹ a + 2 ² b + 2 ³ c = 2a + 4b + 8c

S ₃ = 3 ¹ a + 3 ² b + 3 ³ c = 3a + 9b + 27c

Solving these simultaneous equations to obtain a, b, and c yields the following.

A = 3s _1- (3/2) s ₂ + (1/3) s ₃

B = − (5/2) s ₁ + 2s ₂ − (1/2) s ₃

C = (１ ／) s _1- (１ ／) s ₂ + (１ ／) s ₃

From these three coefficients, a predicted value p ₄ that will be taken by the next sample value s ₄ can be obtained.

P ₄ = 4 ¹ a + 4 ² b + 4 ³ c = 4s ₁ −6s ₂ + 4s ₃

[0043] Thus prediction error x ₄ for the actual value s ₄ is as shown in the following equation, this value is transmitted to the receiving side.

X ₄ = s ₄ −p ₄ = s ₄ −4s ₁ + 6s ₂ −4s ₃

On the receiving side, the transmitted prediction error x
₄ already using s _1, s _2, s _3, which has been restored and, relative value data s ₄ is restored by performing the following calculation.

S ₄ = x ₄ + p ₄ = x ₄ + 4s ₁ -6s ₂ + 4s ₃

In this case as well, of course, it is necessary to finally convert the data into sample value data, and the data can be converted by the same means as described in the description of the power series up to the second order.
Detailed description is omitted.

The case where the power series is expanded to the power series up to the second order and the case where the power series is expanded to the power series up to the third order have been described above. It will be clear from the explanation so far that if this idea is further developed, it can be expanded to higher power series. The higher the order, the more complicated the operation, but the smaller the coefficient value. Therefore, it is not advisable to handle the higher order. In most cases, a sufficient compression ratio will be obtained even up to the second order, and the operation circuit will be extremely simple.

[0049]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, based on the above-described principle, 2
The entirety of a transmitting unit (compressing unit) and a receiving unit (reconstructing unit) configured to perform an operation of expanding a power series up to the next to obtain a prediction error, transmitting the prediction error, and restoring the original sample value 2 will be described with reference to FIG.

In FIG. 2, the upper half shows a transmitting unit, and the lower half shows a receiving unit. 1 is an analog signal input terminal, 2 is an encoder (COD) that performs sampling and encoding, 3, 4, and 5 are shift registers (REG), and 6, 7, 8, and 9 are subtractors (SB).
T), 10 is a tripler (3MLT), 11 is a subtractor (SB
T), 12 is a transmission error control encoder (ECD), 13 is a transmission signal output terminal, 14 is a transmission line, 15 is a reception signal input terminal, 16 is a transmission error control decoder (DEC), and 17 is an adder (ADD), 18, 19, and 20 are shift registers (REG), 21, 22, and 23 are subtractors (SBT), 2
4 is a tripler (3MLT), and 25 is a signal output terminal.

The analog signal applied to the input terminal 1 is
The data is sampled and encoded by an encoder (COD) 2, added to shift registers (REG) 3, 4, and 5 having a three-stage configuration and shifted at sampling timing. This signal is a digital code, and all circuits are operated with this digital code. However, an analog description is easier to understand, so an analog description will be given.

The signals applied to the shift register (REG) are accumulated and shifted one after another in the order of 5, 4 and 3 from the earliest time. In the subtractor (SBT) 6, 1
Since the leading sample value data is subtracted from the sample value data after the time slot, relative value data s ₁ is obtained, and the subtractor (SBT) 7 subtracts the leading sample value data from the sample value data after two time slots. Relative value data s
₂ is obtained. These two signals are applied to a subtractor (SBT) 9 to obtain s ₂ -s ₁ . This signal is then tripled in a tripler (3MLT) 10. In this tripler, it is easier to add the signal shifted by one bit higher and the original signal, as shown in FIG. 3, than to use a multiplier. Since the predicted value p ₃ = 3 (s ₂ −s ₁ ) is obtained by the tripler, the subtracter (SBT) is used in the subtractor (SBT) 11.
BT) From s ₃ obtained at the output of 8, a predicted value p ₃ = 3
By subtracting (s ₂ −s ₁ ), the target prediction error x ₃ is obtained. This signal is added with a redundant bit for transmission error control in an encoder (ECD) 12 and transmitted to a transmission line 14.

On the receiving side, a transmission error control decoder (DEC) 16 corrects a transmission error and removes redundant bits to obtain a correct prediction error signal. Since this transmission error control is not directly related to the subject of the present case, the description is omitted.

First, as an initial state on the receiving side, it is assumed that the contents of the shift registers (REG) 18, 19, 20 and the tripler (3MLT) 24 having a three-stage configuration are all zero. Then, the transmission error control decoder (DEC) 16
The prediction error code which is the output signal of the adder (ADD) 1
7, shift registers (REG) 18, 19, 2
Added to 0. The circuit including the shift register and the three subtractors has exactly the same configuration as the transmission side. Therefore, if the same signal as the sample value data (the output signal of the encoder 2), which is the original signal on the transmission side, is reproduced in the output of the adder 17, the subtractor (SBT) is processed in the same manner as on the transmission side. 23, s ₂ −s ₁ is obtained, and a tripler (3ML)
T) the output of the 24 predicted value _{p 3 = 3 (s 2 -s} 1) is obtained. The transmission error x ₃ sent from the transmitting side is added to the adder (ADD) 17 and at the same time, the predicted value p ₃ =
3 (s ₂ −s ₁ ) and the original sample value data of s ₀ , which is the output of the shift register (REG) 20 which is a reference for obtaining these relative value data, are also added. ) 17 reproduces the same sample value data as the original signal on the transmission side. The operation of each unit when a signal as shown in FIG. 4 is added as specific data will be described. In FIG. 4, a curve indicated by a solid line is an input analog signal, a black circle indicates a sample point, and the sample value is entered. All the registers are reset to zero, and starts the operation from the time slot t _1.

Tables 1 and 2 show the contents of each part for each time slot in this case.

[Table 1]

[0056]

[Table 2]

Table 1 shows the transmitting section, and Table 2 shows the receiving section. In Tables 1 and 2, <2><3>... At the top represent outputs of the circuits having the same numbers in FIG. In the transmitting section, the output of <11> is obtained for the input data of <2>. After the operation starts, a large output appears at t ₁ and t _2.
It can be seen that the t ₃ after the signal enters the One register has become small output level. This prediction error signal is transmitted to the receiving side through the transmission error control encoder 12 and the transmission line 14, and is output to the output of the transmission error control decoder 16 on the receiving side.
A prediction error signal that is transmission data is reproduced. Therefore, the receiving section operates based on the signals shown in <16> of Table 2. The column shown in <20> is the restored data, and it can be seen that the same data as that on the transmitting side has been restored.

As can be seen from the above description of the operation, a code output representing a large amplitude is generated until the system enters a normal operation, but it is necessary to faithfully transmit the code output. Therefore, it is economical to switch the system between a mode in which a code representing this large amplitude can be transmitted for a short time and a subsequent compressed signal transmission mode.

As described above, according to the present invention, by expressing analog data using a power series expansion equation using the sampling time as a variable, the predicted value which will be next taken on the compression side is simply represented by the immediately preceding 2 It is possible to calculate with a more accurate value than when it is on an extension of one sample value. As a result, the prediction error is significantly reduced, and the data can be compressed very efficiently. Therefore, the burden of adding redundant bits for error control is reduced, the data processing capacity is increased,
It leads to faster speed.

Further, by using the second-order power series expansion equation, in addition to the above-described effects, it is possible to simply configure the data calculation method and its arithmetic circuit.

Further, by representing the sample values used for the calculation as relative values, the amount of data in the calculation circuit is reduced, and an excellent effect that the data processing speed can be further increased is achieved. <Other embodiments>

The present invention is not limited to the embodiments described with reference to the above description and the drawings. For example, the following embodiments are also included in the technical scope of the present invention. Various changes can be made without departing from the scope of the present invention.

(1) In the above embodiment, transmission / reception of analog data has been described as an example. However, the present invention is not limited to this.
The prediction error x ₃ calculated and recorded on the recording medium, may then be applied to recording and reproducing system of the data that calculates the original sample value by reading the value. In this case, high-density recording becomes possible.

(2) In the above embodiment, the operation is performed by a so-called hardware logic circuit, so that high-speed processing is possible. However, the present invention is not limited to this, and the operation can be executed by software using a microprocessor.

[Brief description of the drawings]

FIG. 1 is a waveform diagram of an analog signal for explaining the present invention.

FIG. 2 is a block diagram showing an embodiment of the present invention.

FIG. 3 is a circuit configuration diagram for explaining a simple configuration method of a tripler in FIG. 2;

FIG. 4 is a waveform chart of an analog signal for explaining an operation based on specific data according to the embodiment;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Signal input terminal 2 ... Sampling, encoder (COD) 3 ... Register (REG) 4 ... Register (REG) 5 ... Register (REG) 6 ... Subtractor (SBT) 7 ... Subtractor (SBT) 8 ... Subtraction Unit (SBT) 9 Subtractor (SBT) 10 Tripler (3MLT) 11 Subtractor (SBT) 12 Transmission error control encoder (ECD) 13 Transmission signal output terminal 14 Transmission line 15 Reception Signal input terminal 16 ... Decoder for transmission error control (DEC) 17 ... Adder (ADD) 18 ... Register (REG) 19 ... Register (REG) 20 ... Register (REG) 21 ... Subtractor (SBT) 22 ... Subtractor (SBT) 23: Subtractor (SBT) 24: Tripler (3MLT) 25: Signal output terminal

Claims (5)

[Claims] 1. A method of sampling and sampling a continuously changing analog signal at a predetermined timing, compressing the sampled signal, and restoring the compressed signal. On the compression side, a power series expansion equation using the sampling timing as a variable is provided. Is used to calculate a predicted value of a sample value that will be taken next based on a plurality of continuous sample values, and calculate an error between the predicted value and an actually obtained sample value. Using the power series expansion formula with the sampling timing as a variable, calculate a predicted value of a sample value to be taken next based on a plurality of sample values sequentially calculated, and calculate the predicted value and the error. And a method for compressing and restoring data of an analog signal, which comprises restoring an original sample value. 2. A method of sampling and sampling a continuously changing analog signal at a predetermined timing, compressing the sampled signal, and restoring the compressed signal. On the compression side, a quadratic power using the sampling timing as a variable is used. by using a power series expansion formula, calculated by two sample values s _n-2, s _n-1 in will based take to the next sample value s _n predicted value p _n the following equation of a continuous (1) the error x _n and the predicted value p _n actually obtained sample value s _n is calculated by the following equation (2), the recovery side, said second-order power series with the sampling timing as a variable with expansions, calculated by sequentially calculated two sample values s _n-2, s will take based on the next to _n-1 sampled values s _n the equation predicted value p _n (1) with, from its predicted value p _n the prediction error x _n, by the following equation (3) Data compression and decompression method of the analog signal, characterized in that to restore the sample value s _n and. _{p n = 3 (s n-} 1 -s n-2) ... (1) x n = s n -3 (s n-1 -s n-2) ... (2) s n = x n +3 (s n ₋₁ −s _n−2 ) (3) 3. A method according to claim 1, wherein said sample value is obtained as a relative value with the first sample value being 0.
How to restore. 4. A data compression device for sampling and sampling a continuously changing analog signal at a predetermined timing, and compressing the sampled signal, using a power series expansion formula with the sampling timing as a variable. Predictive value calculating means for calculating a predicted value of a sample value to be taken next based on a plurality of continuous sample values; and error calculating means for calculating an error between the predicted value and an actually obtained sample value. An analog signal data compression device comprising: 5. A data restoration device for restoring compressed data of a sample value obtained by sampling a continuously changing analog signal at a predetermined timing to an original sample value, wherein the sampling timing is a variable A predicted value calculating means for calculating a predicted value of a sample value to be taken next based on a plurality of sample values sequentially calculated, using a power series expansion formula, and the predicted value and the calculated value calculated on the compression side. An analog signal data restoration apparatus comprising: a sample value restoration unit for restoring an original sample value from an error between a predicted value and an actually obtained sample value.