JPS6342986B2 - - Google Patents

Abstract

PURPOSE:To perform the coding in an extremely simple and effective way, by giving an inverting process to each element between the signal data having a weak correlation based on the rules different from each other and obtaining the general sum of the elements to compress the data. CONSTITUTION:The aural signal supplied through a microphone is fed to an A/D converter 2 through an LPF1 to be sampled and then fed to a differential circuit 3 in the form of binary digital data. The data of each one second of the signal data series are divided into data trains A, B, C and D. These data trains are led to an adder/subtractor circuit 4 to receive the inverting process based on each rule defined according to each train and then supplied to other signal data trains. The sum signal data train S comprising the general sum data obtained through the addition and subtraction carried out over the data trains A-D is stored in a storage device 6 after compression by a storage device 5 and the circuit 4.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an encoding method that can effectively compress data that has a strong correlation between adjacent sample points such as audio signals and video signals.

Generally, when audio signals are sampled, digitized, and then recorded, reproduced, or transmitted, it is necessary to quantize them with an accuracy of about 8 bits per sampling at a sampling cycle of 8 KHz. This requires a detailed information processing capacity of 64Kbits per second. Therefore, in order to save the capacity of the recording device and narrow the bandwidth of the communication line, various data compression methods have been studied. For example, the PARCOR method, which extracts feature parameters from an information signal, compresses the data, and synthesizes the parameters to reproduce and restore the information signal is attracting attention. This method has the advantage of being able to compress the data rate to about 2 to 4 Kbits per second, but on the other hand, it requires special signal processing to extract and synthesize feature parameters, making the hardware configuration considerably complex. It was hot. In addition, when seeking naturalness and clarity in the reproduced and restored information, the signal processing becomes extremely complicated and is difficult to realize. Moreover, since such signal processing requires a large number of multiplication calculations, there are problems such as the system configuration becoming quite large-scale.

The present invention has been made in consideration of these circumstances, and its purpose is to process signal data sequences with strong correlation between adjacent sample points such as audio signals and video signals without impairing their naturalness. It is an object of the present invention to provide an encoding method that can effectively compress data and can be realized with a simple hardware configuration.

That is, the present invention focuses on the fact that audio signals, video signals, etc. have a strong correlation between adjacent sample points, have periodicity, and at the same time have selective extraction properties against noise, and exploit these properties. The object of the present invention is to provide an encoding method that can be effectively used to realize a new data compression method that is theoretically completely different from conventional methods.

The outline of the present invention differs from the conventional method of extracting compressed data from a group of neighboring data elements that have strong correlations. This method compresses data by reducing the number of elements itself, and is therefore fundamentally different from conventional methods. When reproducing the data, when focusing on a certain data string, other data strings are regarded as random numbers, and this makes the data string stand out. Restoration takes place. Further, the present invention is characterized in that such signal processing can be easily performed in terms of hardware using only addition and subtraction.

First, the properties of audio signals, video signals, etc. are listed as follows.

(a) The correlation between adjacent sample points is extremely strong.

(b) The signal data string is not random, but has a certain degree of periodicity.

(c) The data polarity of one sample point is often the same as the data polarity of the next sample point.

(d) Even if the restored data is not perfect, that is, even if it contains some random number noise, it will not cause much inconvenience in information reproduction.

In other words, in contrast to electronic computers, which cannot tolerate data errors of at most one bit, in the case of audio signals, even if there are some bit errors, it is not possible to read information from the entire signal. It can be done, so there is no problem. This is supported by the fact that, for example, even in a very noisy environment, humans are able to hear only desired sounds. In other words, when focusing on one desired audio data, the remaining data acts like noise and has no meaning with respect to the desired data. Moreover, if the remaining data is completely converted into white noise, the clarity will be dramatically improved. Therefore, in the method of the present invention, by inverting each data string according to rules determined respectively,
The data is compressed by combining each of the data strings so that they become random numbers. As a result, if we focus only on a certain desired data string during playback, all other data strings act as random noise, so we can obtain only the desired data string and perform data playback. I have to. At this time, even if n pieces of data are added randomly, the amplitude will only be multiplied by √ at the most. We are trying to make this happen.
Furthermore, since data processing only requires inversion processing and summation calculation, hardware configuration can be simplified without requiring a multiplier as in the conventional case.

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic diagram of a signal recording device configured using the encoding method according to the present invention. An audio signal input from a microphone is guided to a low-pass filter 1, and after frequency components of 4 KHz or higher are cut out, the audio signal is input to an A/D converter 2. This A/D converter 2 converts the audio signal to, for example, 32KHz.
The data is sampled at a period of 1, and is output as 8-bit binary digital data. However,
The 8-bit data obtained by the A/D converter 2 is sequentially input to a difference circuit 3, and the difference value between two consecutive sample point data is obtained. Although this differential processing is not necessarily necessary, it emphasizes the high frequency range (treble range) and increases the clarity of the encoded and reconstructed signal.

Now, data for the first second of the signal data series preprocessed in this way is converted to data stream A.
(A ₁ , A ₂ -A ₃₂₀₀₀ ), the data for the next 1 second is divided into data string B (B ₁ , B ₂ -B ₃₂₀₀₀ ), and similarly sequential data strings C and D. Although a column divided into four data streams will be explained here, it is actually preferable to divide the data streams so as to realize a multiplicity of four or more in order to ensure the quality of the signal to be reproduced and restored. Further, since each signal data string divided in this way has a lag of 1 second, it can be considered that there is almost no correlation, that is, it is extremely small.

Each signal data string thus divided is led to an adder/subtractor circuit 4, inverted according to a rule determined corresponding to each column, and then incorporated into another signal data string. That is, the first storage device 5
A signal data string that has been previously inverted according to the above-mentioned rules is written in , and the above-mentioned new signal data string is added to this according to the established rules. This first storage device 5
and the addition/subtraction circuit 4, the sum signal data string S consisting of the sum data added and subtracted over the signal data strings A to D according to a predetermined rule is stored in the second storage device 6.
It is now stored in . Incidentally, a series of operational controls of these parts is performed by a control circuit 7.

Now, the summation processing of the signal data string by the above-mentioned addition/subtraction circuit 4 and first storage device 5 is performed, for example, by the following procedure. The basic signal data string output by the difference circuit 3 is shown as in FIG. 2a. The control circuit 7 uses its built-in counter to generate a carry signal each time it counts, for example, 32,000 elements of the data string, and thereby divides the data string into the aforementioned four elements A, B, C, and D. At the same time, the carry signals are counted, and the signal data string is identified from the counted values 0, 1, 2, and 3. Further, corresponding to this identification signal data string, different rules are followed between each signal data string. Generates inverted/non-inverted inverted control signals. This inverted control signal is, for example, non-inverted (indicating arithmetic control form) with all "1"s for signal data string A, and non-inverted (additional) for signal data string B. This shows a control form in which the and inversion (subtraction) are repeated alternately. Also, for signal data string C,
Addition, addition, subtraction, subtraction is a control form in which addition and subtraction are repeated every second, and addition and subtraction are repeated for every second signal data string D as well as signal data string C. is returned, but the repetition is 1
Each shows a control form in which the timing is shifted. Control information for these additions and subtractions is stored as "+", "-", or as "0" and "1" in the shift register of the control circuit 7 according to the corresponding signal data string. According to these control rules, each element of the signal data string A is first stored in the first storage device 5 as addition data.
written to. After the writing of each element of the signal data string A is completed, the signal data string B is input to the addition/subtraction circuit 4, and these signal elements are sequentially added and subtracted alternately and stored in the first storage device 5. are filled into the elements of the corresponding sampling points stored in . Thereafter, the data will be integrated in the same manner for the signal data strings C and D according to the rules. As a result, each element at the sampling point i of each signal data string is integrated as shown in Figure 2b according to the rules of each data string, as Si=Ai±Bi±Ci±Di. Become. The total sum Si constitutes a sum signal data string as shown in FIG. 2c, and is obtained by compressing the four signal data strings. This sum signal data string S is then written to the second storage device 6, where the data is recorded in a compressed state. In this case,
Only the polarity sign of the sum total value may be extracted as the sum signal data string.

In this way, the sampled data for 4 seconds is 4
After being divided into two signal data strings, data is compressed by concatenation processing according to a predetermined rule, and this process is sequentially repeated.

By the way, each signal data string A, B, C,
D is reproduced and restored as follows. In this case, each element of the data string S is the data string A, B,
It may be considered that the total of the elements of C and D is 10 bits, but it is of course possible to use 1 bit as the polarity sign.
The sum signal data string S generated by the data compression described above is data strings A, B,
It is used four times corresponding to C and D. Now, when reproducing the signal data string A, the control information according to the rules used in the previous consolidation is used in exactly the same way,
This is done by multiplying each element of the sum signal data string S by the polarity sign. Specifically, element Si
This is realized by using the exclusive OR of the control signal and the control signal. Therefore, now when element Si is positive,
If the sign is "+", the data is reproduced as "0", and conversely, if it is "-", the data is reproduced as "1". Further, when the element Si is negative, the data is reproduced as "1" if the sign is "+", and as "0" if it is "-". Similarly, signal data string B,
The data of C and D are also reproduced as shown in FIG. 2d by multiplying the polarity codes according to the rules corresponding to the data strings.

That is, as is clear from the data array shown in FIG. 2b, if we focus on a certain signal data string and multiply that signal data string by the polarity sign used for inversion processing, we can obtain the desired signal data. Each element of the data string is shown as a positive data string, whereas the other data strings have their polarities reversed within the data string, albeit periodically. Therefore, due to this polarity reversal, the data component does not lose the information that it should originally have, and can be considered to be noise. As a result, the only data string that shows information in its original form is the signal data string of interest, which stands out compared to other data strings. Therefore, by inverting the sum signal data string, it becomes possible to obtain information that strongly reflects the information of the data string used in the inversion process, and the data is then restored and reproduced.

In actual data reproduction, if the information elements obtained by the above-mentioned restoration process are further averaged between adjacent sample points, information reproduction with higher precision, that is, with a lower noise level, can be achieved. It becomes possible.

As described above, according to this method, the properties of audio signals, video signals, etc. are actively utilized, and each element between signal data strings with weak correlation is inverted according to mutually different rules. By calculating the sum and performing data compression, the purpose can be achieved extremely easily and effectively. Moreover, since the data is reproduced by converting other data strings into noise and extracting features, it is possible to obtain information that strongly reflects the original information, which is extremely effective.

Incidentally, the inversion processing rule used for such data compression can generally be easily given as a column of each element of an orthogonal matrix. For example, it may be given as element Hi.j of Hadamard matrix H. In other words,
When n signal data strings are divided and set, each element of the n-th order Hadamard matrix may be used and read out cyclically to be used as a control signal for the inversion process. For example, in the case of the 8th order: ＋＋＋＋＋＋＋ ＋−＋−＋−＋− ＋＋−−＋＋−− ＋−−＋＋−−＋ ＋＋＋＋−−−− ＋−＋−−＋−＋ ＋＋−−−−＋＋ ＋− −+−++− Each column element of the Hadamard matrix may be used as inversion processing control information corresponding to each signal data string. Also
In the case of order 2 ^m+1 , the Hadamard matrix of order 2 ^m is expressed as H(2 ^m )
When [H(2 ^m+1 )] = H(2 ^m ) H(2 ^m ) H(2 ^m ) - H(2 ^m ), each element shown in the matrix may be used.

As mentioned above, as the method of the present invention has been clarified through the embodiments, according to the present invention, the required data compression can basically be performed only by addition/subtraction processing. Furthermore, when data is compressed in this manner and then restored and reproduced, the original signal is strongly reflected and there is sufficient reproduction capability.

Now, let's take the case of calculating S with 11 bits as an example.
Since audio signals basically show gradual waveform changes, slow signal changes (Roman letters)
and a rapidly changing signal portion (Greek letters), for example, as shown as A=a ₁ +α ₁ (a ₁ ≫ α ₁ ). When the reproduced element is averaged with the reproduced element of the adjacent sampling point and this is reproduced data A', it is expressed as A'=1/4 (S ₁ +S ₂ +S ₃ +S ₄ ). However, the addition and subtraction codes follow the predetermined rules described above. Therefore, focusing on the right side of this equation, it is transformed as follows.

1/4 (S ₁ + S ₂ + S ₃ + S ₄ ) = 1/4 (A ₁ + B ₁ + C ₁ + D ₁
)+1/4(A2 _{- B2} + _C2 - _D2 )+1/4( _A3 + _B3 - _C3 - _D3 )+1/4( _A4 + _B4 - _C4
+ _D4 )=1/4( _A1 + _A2 + _A3 + _A4 )+1/4(B1 _{- B2} + _B3 - _B4 )+1/4( _C1 + _C2 - _C3
-C ₄ ) + 1/4 (D ₁ -D ₂ -D ₃ +D ₄ ) = 1/4 (4a ₁ + α ₁ + α ₂ + α ₃ + α ₄ ) + 1/4 (
β ₁ −β ₂ +β ₃ −β ₄ ) +1/4(γ ₁ +γ ₂ −γ ₃ −γ ₄ )+1/4(δ ₁ −
δ ₂ −δ ₃ +δ ₄ )=a ₁ +α ₁ +(error) This error Δ can be almost ignored because it is a small alternating current component that is randomly added or subtracted. Therefore, highly accurate reproduction data can be obtained here.

On the other hand, the processing time required to execute this method is three operations for one sampled value: (a) reading from the storage device, (b) addition/subtraction processing, and (c) writing to the storage device. Generally, each operation can be performed in 1 μsec. Furthermore, since the sampling value period is 30 μsec, it is possible to perform sufficient real-time processing. In this regard, in the conventional PARCOR method, for example,
Several tens of multiplications and additions must be performed within 100 μsec. In comparison, it can be seen that the present method allows for an extremely simple circuit configuration and has an extremely high degree of data compression.

Note that the present invention is not limited to the above embodiments. In the embodiment, data compression is performed by configuring dedicated hardware, but this may also be performed by software using an electronic computer. Further, it is not necessarily necessary to store the signal data string in a storage device and perform data compression processing. For example, four independent audio signals are input as shown in FIG.
After passing through the polarity reversing circuits 12a, 12b, 12c, which have mutually different control regulations,
12d, and input its output to the adder circuit 13 to multiplex and synthesize four independent signals, that is, four channels of signals. Further, the above-mentioned inversion processing can be applied not only to polarity inversion but also to carrier wave amplitude modulation. In this way, multiplex communication of radio waves becomes possible, and in this case, one amplitude S may be made to correspond to each carrier wave, or one amplitude S may be made to correspond to each plurality of carrier waves. Moreover, in this way, compared to the conventional method, there is no need to set a guard frequency band, and extremely high multiplicity can be achieved. In the case of video signals, multiplexing may be performed between separate scanning lines, between different frames, or between completely separate image signals.
In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of a signal recording device configured using an embodiment of the present invention, and FIGS. 2 a to d
3 is a diagram for explaining the operation of this system, and FIG. 3 is a diagram showing the configuration of another embodiment. 1...Low pass filter, 2...A/D converter, 3...Differential circuit, 4...Addition/subtraction circuit, 5, 6
...Storage device, 7...Control circuit, 11a, 11
b, 11c, 11d...A/D converter, 12a,
12b, 12c, 12d...Polarity inversion circuit, 13
... Addition circuit.

Claims (1)

[Claims] 1. Means for sampling a continuous input signal and dividing it at predetermined time intervals to generate a plurality of sampled signal data strings, and for each signal element of these signal data strings for each column. means for performing inversion/non-inversion processing according to elements of different columns of the Hadamard matrix; means for calculating the sum of each row of the signal data strings subjected to the inversion/non-inversion processing to generate a sum signal data string; An output means for outputting a signal data string, an input means for inputting the sum signal data string outputted from this output means, and each signal element of the sum signal data string input from this input means and the inverted/non-inverted and reproduction means for multiplying each row by the signal element of each column of the Hadamard matrix used for processing, extracting the signal data string from the sum signal data string and reproducing the data. Characteristic encoding method. 2. The encoding method according to claim 1, wherein the sum signal data string is constituted by a polarity code of each summation element. 3. Means for sampling continuous input signals and dividing them at predetermined time intervals to generate a plurality of sampled signal data strings, and for each signal element of these signal data strings to be a product of two arbitrary columns for each column. means for performing inversion/non-inversion processing according to elements in different columns of a matrix such that the sum of the sums is approximately zero; output means for outputting the sum signal data string; input means for receiving the sum signal data string output from the output means; and each signal element of the sum signal data string input from the input means. is multiplied row by row by the signal element of each column of the matrix where the sum of the products of any two columns used in the above inversion/non-inversion processing is approximately zero, and the above signal data string is obtained from the sum signal data string. and reproduction means for extracting the data and reproducing the data. 4. The encoding method according to claim 3, wherein the sum signal data string is constituted by a polarity code of each summation element.