JPS61500998A - System and method of data compression for audio signals - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] This is Charles Niss Quiver (Charl@sB-W@aV@r) US patent application cylinder 202. Hiro! No. 7, “Methods and Equipment for Digital Data Compression” Place J (”M@thodand Apparatus for Tffglt* I Data Compr@5slon’) (filed on July 3, 1910) ) is a partial continuation application.

After converting an analog signal to a digital signal for recording or transmission to a remote location , means for performing pressure mF wave and Huffman encoding of a signal, and a Huffman decoder , a digital restoration F-wave generator, and decoding, converting the F-wave and nine-day zotal signal to For systems containing playback or reception means including means for converting back to analog format. The above-mentioned U.S. Patent Application No. 202.4tj-7 and I.E.T. Transactions on Biomedical Engineering magazine, ・The paper “Prediction or Compensation” written by E. Latchman and H. Gui Bigberger Compression and entropy encoding of ECG by J (U, E, Rssttim mnn and H-V.

Pipb@rger+@Campress1on of the ECG by Predietloz+or! ! 1terpolation and En tropy E+eoding” * IEETransactions 6! l　Biemedreal　Engineerfng, Vol　・BKE-21s , A/ /, pp-6/3-Otsu, 2j, Nav, /ri7ri) K mentioned et al. It is. Similar systems are available at K.L. Ripley and J.R.Co. [Transient electrocardiograph - A computer for capturing deku] Data system J (K, L, play and J, R.

Cox Jr+, -mu Computer System for Capturi ngTransient El@etroeardi*grapbic Dat a”r　Pre.

Comptlt, Cardiel, pp, 'A jter≠≠! ,/ri7otsu) It is stated in According to the present invention, analog signals such as music, electrocardiographs, electroencephalograms, etc. The average pit speed of the digitally converted audio signal is sufficiently reduced, resulting in low quality digital transmission over a transmission line, or by using a relatively small amount of recording media and a known digital Significant amounts of signals can be recorded and played back using digital recording and playback techniques.

Summary of the invention Audio signals such as music to be transmitted or recorded are converted using analog/risota A/conversion means. is converted into digital format. Next, this digital signal is 2 isotally compressed p A compressed digital signal is created. compressed audio signal For encoding, an encoder such as a truncated Huffman (fIaffman) encoder is used. given to.

The digital output from the encoder can be recorded remotely using digital tone means. One or both of the transmissions to the receiving location may take place. Playback device or receiving stage At No. 1, the encoded signal is decoded by a decoder, and this decoded signal is is applied to the file decompressor Fi.

The output of the expanded F wave generator is converted to analog form by digital analog F converter means. is converted and the audio signal is played back. smoke volume or transmitted digital signal word A Deinotal compression-expansion waver combination is used to minimize the average pit length. . Deinotal pressure (The transfer function of the JP wave device is at zero degrees on the unit circle of two planes. The zero point is up, r isotal extension P sinker's cloud 2! The function is a 2,000-sided unit circle. There is a 1 at zero degrees inside the circle. Compressed F-wave generator calculations are performed without truncation or rounding. In contrast, truncation or rounding is used in the operations of the decompressor F#. zero In addition to the zero point located in degrees, the transfer function of the compressed F-wave device is located at +L1. Has a zero point at Hiro10°±6OZ±70@, ±/20°, and/or l♂O0 be able to. As a corresponding extended F-wave device, pressure i is placed on or inside the unit circle. An aP wave device having a pole adjacent to the zero point position is used. The resulting image The frequency response of the stem for analog No. 1 is a high-pass type, with 7 or more small may contain high-frequency notches. For music signals, the resulting F-wave The device has a low frequency cutoff between OHz and /yo-! Hz to 2 It can pass audio frequency signals ranging from 0.000 Fix.

If the transfer function of the expansion Pi unit includes colors on the unit circle of two planes, then the compression-expansion P The i-device combination becomes unstable. For such configurations, the Huffman encoder Uses a false check code in the output transfer and eliminates the false check code during transfer to the Huffman decoder. It is possible to provide error detection means for detecting. This kind of digital signal transfer An erroneous signal is generated in response to error detection during transmission.

This error signal is applied to the extended F-wave generator, which temporarily moves its pole inside the unit circle. . This allows the system to recover from the signal error.

Even in a system equipped with a stable compression-expansion F-wave generator combination, the above error signal detection) and inside the poles of the transfer function in two planes of the extended F-wave generator in response to error detection. can be used to speed up recovery from signal errors.

Error checking code in systems containing unstable compression-expansion F-wave generator combinations Instead of using erroneous signal detection means, the actual signal value at once from the A/l) converter is The data is periodically transferred to the restoration door breaker, and when an error occurs, the restoration door breaker periodically sends the message ``again.'' The system may be operated in such a way that it is set to "initial settings." For music signal compression, For example, it is appropriate to transmit the actual signal value every 6 to 1 milliseconds. . The number of consecutive actual signal values that must be transmitted periodically is It depends on the number. The actual number of signal values sent out is equal to the above order.

Special edition Showa 61-500998 (5) First, Figure 1A shows the digital recording-transmission section of the data compression system. . This includes analog audio signals/(t)tanalyzer for converting to digital format. A log-to-digital converter (AD converter) 20 is included. analog-r iji The nth temple from the barrel converter 20 is denoted by f. For the second bad guy Analogue signal = 1 is shown, which is the input of analog-to-digital converter 20. It is power. For illustration purposes, the audio input signal has a frequency of approximately /j　Tlz to 20,0 It is assumed that the music signal is in the range of 00Hz. The analogue stage shown in Figure 2B The output format of the digital converter has equal word lengths of 7”A/fI, -, to ' n+i is included. Analog-to-digital converter 20 connects timing line 26 to The signal set by the control signal given from the timing control device 21A via the operating at sampling speed. Timing control device, as used here Line 2B from 21A represents multiple timing circuit outputs, of which 7 or more Outputs are provided to system components for proper system timing and control. It will be done. It is also controlled by signals from various other components of the system, so Input is provided to the timing controller via 2g. The A/D converter 20 is normally operates at a constant sampling rate and delivers a constant word length output.

The ~Φ converter operates with a sampling rate of ≠ kHz and a word length of ≠ bits. child The values are for illustrative purposes only and are not intended to limit the invention.

) The output of the JD converter 20 is converted to the deisotal pressure MP via the digital/I/wave device λ3. wave generator 30. Digital signal converter 23 Yes Digital from converter 20 Signal entropy is achieved by de-encoding the high frequency part of the audio frequency signal. Reduce peas. Figure 2A shows the frequency characteristics of the 7-wave generator 23 and the regenerative reception of the system. Deisotal p-wave device 7 included in Shinbu! shows the frequency characteristics of For simplicity, Together with the deinotal output of the door wave device 23, it is expressed as r isotal input tf.

〜勺EN#) Input of converter 20 of analog F-wave converter with frequency characteristics similar to iC It is clear that it may be provided at -11.

Digital pressure, for the current monthly edition! The P-wave device 30 includes an evaluator 32 and a subtraction means 3I. Contains A. The evaluator 32 occurs before and after the sample JL/f to be evaluated. There is an actual sun. A difference signal is obtained from the pressure MF wave generator 3θ. This difference signal Δ is actually The difference between the actual signal input f and the evaluation signal value is n! I! I Ah, by subtracting the evaluation value from the actual value using subtraction means 3≠ as shown in the following formula. can get.

In the graphical signal representation of the pressure waver output shown in FIG. 2C, the difference signal Δ, Δ, Δ, ... Δn+i is n n+1 B+2 It is shown. According to seven features of the invention, the arithmetic of the digital pressure mP wave generator 30 The operations are performed without truncation or rounding, whereas the corresponding digital expansion described below Arithmetic operations on length or restoration Fe, containers are performed by truncating or rounding. Figure 2C As shown in Figure 1, the output of the 8-channel waveform is a compressed signal with a length of 7r bits without truncation. be.

It will be understood that this is not something to be ignored. Compression door waver output Δ is the actual input f The difference between and its evaluation value; In Another f! that is not a function of Other pressure +2fa devices with 'C' may also be used. "difference The term "signal value Δ" can also be used to indicate the output of other suitable compression door breakers. Ru.

The compressed signal value Δ is truncated to encode it via switch 3J''k. It is given to an encoder Hiro0 using Huffman code. Huffman encoding is "Method and Apparatus of Pending U.S. Patent Appearance of Luz Niss Quiver" (disclosed) (U, S, Pat@atmuppHcation S@real 4207,7 2r, Charles S.

Weav*r+ @Method and Apparatus for di g1talH+xffman Enc-oding -) (/210//month (filed on the 77th, applicant is the same as the present invention). The above U.S. Patent No. B θ7,72♂ The entire disclosure is cited in the non-PA specification. Briefly stated, He7man encoding method Then, since the compressing p-wave generator reduces the entropy of the signal output Δ, it is compared to the input signal. By using the fact that the number of bits of a Huffman encoded signal can be reduced by There is. Seven code words were assigned to the rarely occurring differential calls, and the actual differential It is given as a label for the signal value Δ. In the first diagram, the output of encoder HiroO is b( Δ! 1). At D in Figure 2, the value h(Δ, ri, h(Δniya,) etc. is Δ.

It represents a coded cage such as ΔtI+1. Δ. The most frequently occurring value of ( Here zero) is encoded using the shortest code word. truncated hidden hafma The code is disclosed in U.S. Patent Application No. 207.721, which is a simple It is easily realized using an encoder and a decoder. The output of encoder lAo most often contains Regarding the value of ΔU that occurs; contains a combination of the code word label and the actual value of the compressed signal ΔU. ing. For the sake of explanation, if the compressed signal value exceeds ±3, it is actually used as the encoder output. The compressed signal Δ and the code P word label are obtained. Figure 3 shows several encodings. The encoded value for Δ11+2 has a label and The actually compressed signal Δ1□ is included. Here, ΔU+2 rarely occurs It is a compressed signal value, ie a value outside the range of ±3.

The encoded signal from encoder≠0 is either recorded or transmitted to a remote receiver. Either or both may be done. In the case of recording, the encoder output is routed through the switch. is connected to a recording device θ, and the encoded difference signal h(Δ) is recorded. Swish When CH≠r is switched to the opposite side, represented by the dotted line, the encoder output is switched to the buffer Digital modem from memory λ! ≠ and sent to transmission #5 B . Some embodiments of the invention provide encoded data for recording and/or transmission. A check bit is created along with the compressed signal h(Δ).

In some embodiments of the invention, the isotal input signal f is sometimes switched to the switching means 3! to the input of the Huffman encoder. This signal corresponds to the corresponding digital signal described below. O serves to initialize or reset the F-wave device. Recording device for the first map! The encoded digital signal, such as the signal recorded at - Reproduced using the system shown in Figure 8. This system has a playback device It is included. The recorded encoded digital signal from the playback device 60 is truncated 7 7 to a decoder 66 for decoding the encoded signal via a switch 61IL. It will be done. The decoder 6 converts the Huffman code word into a compressed signal Δ. will be replaced. The Huffman code word contains the actual compressed signal with a label. If so, the label is removed from it and the actual unlabeled compressed signal is decoded. given to the device output. Encoding means and decoding means that can be used in the present invention is described in detail in the above-mentioned pending U.S. Patent Application No. 207.72♂. . Encoding and decoding will be explained in more detail in "Encoding-Decoding" below.

The compressed signal Δ from the decoder 66 is decompressed or decompressed via a 72ft buffer memory. The signal is given to Nagatoha device 70.

The output signal of the decoder is produced at a slightly varying rate and includes a decoder memory 72. It is. The restoration Ptt unit 70 transforms the compressed signal Δ into a sample signal of equal length /( out). This signal f (o-t) is the input sample of the compressed p-wave generator 30. The value becomes very close to the signal f. As mentioned above, the seven q!s of the present invention! According to the f-sign Compressed P-waves without truncation and stretched P-waves with truncation are included. As shown in Figure 2F , the output of the truncated restoring p-wave generator (float).

f! , +'(o5t), etc. are composed of 2 bit words. Don't cut everything off In this case, the restoration F-wave generator must handle word lengths of about 3 bits to 0 bits. There will be no. For consumer products, this word length is currently a reasonable cost. It cannot be achieved with. Perform compressed P wave without truncation, expand with truncation Appropriate data compression with minimum distortion can be achieved using a compression-expansion p-wave unit combination that performs Tonami. The reason for this will become clear later.

Digital-analog converter (D/A converter) 7≠ is for reproducing analog signals , converts the signal output from the digital restorer 70 into analog form. change to---n exchange. The digital door detector 7 emphasizes the high frequency components of the signal output. It is provided between the main transducer output and the D/A converter. Frequency characteristics of F wave generator 75 is the second AI8! It is shown adjacent to the frequency characteristics of the uK% input cathode waver 23. simple For simplicity, the same symbol f, (o-t) is used for the input and output of the p-wave generator 7j. . A similar frequency characteristic is applied to the output side of the D/A converter in addition to the digital door transducer. It is clear that an analog P device having the same function may also be provided. Reception timing control The device 7B connects various receivers via line 7 to ensure correct timing of the receiving operation. provides timing signals to the components. Also, the line termination from various components of the receiver. The control signal is given to the device 7B via 0.

In the case of transmission without recording, the encoded signal is →54 (from Figure 1A to Figure 1B) to the receiver's digital modem R2. Modem output is buffered ·memory? ≠, and the buffer memory output is at the dashed line position. is provided to the decoder 66 via the receive.

Quantization of analog music signals Consider a music analog signal as an input to the current system of the present invention. This is an explanation This is not intended to limit the invention.

of a binary X-bit length sample converted from analog to 2 FI2bee is expressed by the following formula.

2! H(J-Σ-P, log2P1 (2) 1=1 However, there are 28 possible sample values, and the Eth value of PK occurs. It's a probability. Assume that it is a scale of the quantization level, and for simplicity, give the first value. Assume that the first quantization obtained is q(l-/) to ql.

At this time, as shown in Figure 1A, the signal before analog-to-digital conversion is q ( i − / ) to qt. The shaded area in Figure 1 is the signal/(i) The probability of entering the first quantization is given.

Furthermore, the magnitude of the quantization level, 9, is small compared to the standard deviation σ of the analog music signal. Assume that If we increase the word length of the converter by l bits, the quantization size becomes is halved, and two quantization bins are formed from the original bin, as shown by the fourth break in the figure. be done. When the distance is small, the areas on both sides of the rotational failure Ps are almost equal, so /(1) is about 10K in two new boxes? , /2. therefore , the entropy of a word of n+/e length is approximately It changes as follows.

The entropy of a word of length (X+/) bits is:

From the above, as the bit length increases, the increase in entropy is added to the word length. It can be seen that each pit that is generated converges to l pit. The first quantization bin is almost zero. The argument is a little more complicated when the center is the center (the usual case), but the result is the same. Ru.

P, tentropy is estimated by numerical integration at various ratios of σ and q for the Gacks distribution. From the following table showing the calculated entropy increase at various ratios of σ/q, ・Every time q is halved, that is, the word length is increased, the entropy increases to l pits. You can see that it's very close.

table / Quantization rhino, when e is reduced Calculated entropy increase of 32 yo 73 Pressure mPi wave of quantized music signal A 7-man encoder such as an encoder 4tO to which the output of the pressure waveform generator 30 is given. The average word length of is constrained as follows.

H (q) < average word length < HCq) + / (5) To realize a compression door waver If the coefficients of the equation(s) used for The child level becomes smaller. That is, the minimum difference between possible output values becomes smaller. , H(q) becomes larger゛. For example, to realize the conversion of a compressed F wave generator, the following two Consider the formula for two compressors.

Δ = / -2f + + / - (6) n n fi1112 and +! , 2-2"-m+1f, 2+2-2rnf, 2 (7) where m is a positive integer It is.

The 2 transformation of equation (6) has two zeros at (/, 0) on the 2,000 plane, and equation (7) has two zeros at (/- There are two zeros at 2″″m, 0).

Δ in equation (7) has values arranged at intervals of 2-2Bq. When m is large , the σ from both F-wave generators are almost equal, but the ratio of σ/quantization level is 2-2m The only difference is the coefficient of . Therefore, the binary trol in equation (7) is the entry in equation (6). There will be approximately λm bits more than the ropy. And after H77 encoding, the average bit length is Approximately 2m bits longer.

Multiplying the right side of equation (7) by 221 returns the quantization level to q, but the standard deviation becomes 2. The general form of the differential equation for the compressor Pa is as follows.

Here J is a constant. If ai can be expressed as a finite-length binary number, this becomes It can be expressed as the formula.

where, = 0 or +/, j can have a positive or negative value.

If j is negative and bi is nonzero, then f! l-1t-shift to the right and add It means something. The least significant end of the arithmetic word K j a pit must be added. No. Here, je is the most negative j for which btj is non-zero.

The 2 conversion of equation (6) is as follows.

G(z)=(/-xz-'+z-2)=(/-z-')2(10This is shown in Figure 5. As shown, it can be represented by two zero points (/, 0) on two planes. All zero points r Frequency characteristics at f of an innotal P-wave device [for example, equation (6) or equation (7)] are the distance from the point exp (12r/r) to each zero point and the gain constant [in equation (10, / equal to]. Here f is the frequency, and the time between the samples is It is. Therefore, the frequency characteristics of equation (to) are as follows.

R(10) = 42Ql) If there are n zeros at (/, 0), R (/, ) = d'' αa These pressure MF wavers reduce entropy for the following reasons. sampling If the speed is Hirohiro x lOs samples/sec, the point on the unit circle (-/, (7) is 22 Corresponds to a frequency of kHz. The centroid of the music spectrum is usually /yoJj/kHz It's small. Therefore, most spectral points are close to (/　,O) on the unit circle corresponds to the point. The value of d (and aTl) becomes progressively smaller. set σ(Here, the integral of the spectrum value time d′ as IWff of θ=Jπ/n is input smaller than the variance of kus(cut A/(K≦/). Decrease in variance is a decrease in entropy. means.

When θ>60° (/=7.33 kHz), the value of d is greater than /. However, Therefore, for a wave generator with all zeros at (7,0), from 7.33 kHz The upper spectral component is amplified by 4n. nt - 7 if larger than a certain value; 7,33kHz is less than the amount by which the total energy below 3jkHz is attenuated. There is a value such that the total energy above is amplified by a greater amount. The value of this n is Minimize output variance and entropy. When -/, the input and output q are equal It is from. This is expressed as follows.

G(Z) = K (/-z-') n where town is the constant used in equation (8) Yes, K is a gain constant. a, is an integer that can be expanded by ° without negative j [Equation (9) ], so q does not become smaller. When K is a power of 2, the value of K Note that the ratio σ/q or the entropy does not change even if It should be done. This is because the input words are only shifted. in this way and minimize the bitropy t-for K=/! 1 also applies to other K The minimum entropy is obtained when K is a power of 2.

The other two zero positions on the unit circle that do not reduce q are at (-/, 0), and ( 0.7) and (a, -i) (complex pair). 2 conversion is as follows.

G(Z)=(/+Z-')" (14) CC-/, 0) with n zero points] and Beauty G(Z)=(/−Z−”)”(1!9 [(θ,/) has n zero points and 0.− IKn zero points].

The other two complex pair positions that do not change q have the following transformations.

G(Z) = (/ −Z−’ + Z−2)” QG (Is this the light on the unit circle? n zero points are placed at the angle +6Oo and n zero points are placed at the angle -O0) and and G(Z) = (/+Z +Z)” (17) (This is n +” at +/10° J■zero point, -/21 add n zero points to 20'').The above is qt-smaller. It is only the zero point position that falls inside the unit circle or the same as the unit circle.

There is nothing outside the unit circle that would be a satisfactory reconstructed F-wave generator.

Another important zero point position for the music data pressure line is Shihiro/, 4t10 complex on the unit circle. They are a pair. The conversion is as follows.

c(z) = / - /, jZ"' + Z-2 (LLG This angle is θ kHz 9 is divided by 2 for each complex pair.

According to the / aspect of the invention, a compression waver is used, and the two conversions are (/, 0) and and the above complex counterpoint t on the unit circle °, ±yo”.

±/　20'' and /♂00).The above zero point position The location is shown in No. 6-. As mentioned above, these zero positions on the unit circle are tropy is minimized and the combination of zero positions used is Depends on the spectrum.

For example, it is possible to reduce the output dispersion due to high frequencies (e.g. 3 to 1kHz). Therefore, the zero point can be placed at a point of ±60°. For this /, there are many numbers for 0 Many zero points can be used. Figure 7 shows the frequency characteristics of three different compression p-wave generators. It is shown. The zero points of the two conversions of these p-wave generators have the units θ0 and θ0, respectively. Exists at 08 and ±O00, and O’m 20°, and 20” . The ability to minimize entropy using the location of the zero points allows for a wide range of frequency characteristics. It is possible to design a pressure MP wave device having the following.

Of course, there is a limit when increasing the number of zeros.

The amount of calculation required is directly proportional to the number of zero points. And Tonami's arithmetic word length has zero points added. It increases by at least / every time it is added. Also, recovery from pit errors during restoration becomes longer as the number of zeros increases.

After the transfer function of the compressed FfIL generator is selected, the resulting entropy is expressed as can be evaluated. The music spectrum 5(1) is measured along the unit circle ( /, O) to (-7,0) , /"l G(Z) l" Sz(e"") ax tL9 is integrated. It will be done. here Sz (””’) = 8 (1) Dechiru. The square root of the integral is σ of the pressure JliF waver output.

The table / can be used to evaluate H(q).

Compression Fi device To realize the conversion of the above pressure IJP wave device! Program P Digital Combier Although it is clear that standard two-digital techniques such as the use of data can be used, Equation (6) The f-lock f-lock of the 92-order digital compressed F-wave generator suitable for use in order to realize this is shown in Figure 2. is shown. Reference is made to FIG. 2 below. The illustrated pressure aF transducer includes a series of 102, 104t, and 10ft registers are included. Is it a φ converter? These continuous sample signals are transmitted to these gate registers via the door transducer 23. It will be stepped in. The figure shows registers 102, 10IA, and Sump Nfn for 10 Otsu each.

/ It is drawn as if it contains /. /≠n-1,,-2 For samples of bits, the /≠pit register is used. out of register The power converts the sample signal into an arithmetic logic unit (ALU-ir1thm*tie and Digital master for selectively connecting to loglc unlt) / /θ It is connected to multiplexer 101. Multigrek? 10♂ and ALU / /0 is controlled by a timing controller @2tA.

As previously mentioned in the explanation of FIG. 1A, the digital pressure JP wave device 30 The inclusion of the vessel 32 results in a force t. The output of this evaluator 31 is the sample f to be evaluated. n, many of them send out all of the next output.

Here, the coefficients 凰 and a2 are selected so that the root mean square error of the difference Δyu is at most 74%. be done. As shown in the previous equation (1), Δ=/−f. J = 62 = For /, &1, formula (1) and rl n n +21), the following formula is obtained.

Δ, = /, +, -21n+f, -1elf) (Here, equation (6) and @bu are I strongly believe that they are equivalent. ) samples on both sides of f, i.e.! Not himself f ! 1-1a tt It is made using fゎ+1. Under the control of device 2hiro, word f! l-1 and' Is +1 to 1 a multiplexer lO? is transferred to ALU / / 0 through calculated. Next, the actual sample f is passed through the multiplexer 70♂ to the ALU/ / Moved into 0 and multiplied by 2. The power of 'X' is simply each bit t - the most significant bit. Shifts toward the cut.

, -f is subtracted from the actual value multiplied by 2, ! In A compressed signal value Δ is obtained at the output of ALU / / 0, which is sent to the encoder tlLO. Given. Operations within LU / / 0 result in truncation or rounding error χW This is done with a sufficiently long word length so as not to be affected by e. Data pressure by the above device It can be seen that the reduction includes evaluation of Nangle values by interpolation.

Huffman encoding and decoding As mentioned above, we use this method to encode and then decode the 3-aP waveform output. Regarding Huffman encoding/decoding means suitable for Permission application number! No. 07,72r, “Method for Digital Huffman Coding” and Ura M” (examination on /9♂O year//month/7th). Seven application contents are books Cited in the specification.

Figure 2 shows the 7-man code to the truncated version.

This is for illustrative purposes only and is not intended to limit the invention in any way. For happiness and misfortune 3, the compressed signal Δ, the knee P word for that signal, and the length of the code word. , as well as a table of the relative probability of occurrence of the compressed signal. most frequently occurring Compressed signal Δ! I (here ±3 or something in between) assigns the code word. It will be done.

The probability that Δ is the value assigned to the code word fc is high; for example, a9r Dechiru. These compressed signals are assigned code words of different lengths.

The most frequently occurring compressed signals are assigned the shortest code words. at the table is the most frequently occurring compressed signal Δ = OK The shortest code word is assigned (the longest food word is divided into the least frequently occurring compressed signal Δunity 3) Can be guessed. All other compressed signals outside the ±3 range are represented as 7 in the table. It is.

These compressed signals are assigned a code word, and this code word is As explained with reference to FIGS. 2 and 3, the curve for the actual compressed signal value ΔU is contains a bell, which in turn records and transmits via a communications link a /-7 machine. is sent to the decoder. The system uses the illustrated truncated Huffman code. It is clear that this is not limited to. The code is added to the additional compressed signal Δ. A code word may be assigned, or other code words may be assigned.

Restored F wave Relationship between σ-1 and the stability of the restored F-wave device Deinotal Sanjiji wave device 30. give to In order to accurately restore the digital music signal, the transfer function of the restoration p-wave generator 70 is It must be the opposite of the transfer function of the compressed P-wave device 30. (For accurate restoration The other two requirements are to avoid overflow errors and under-70 errors in the calculation of the door waver. There is no truncation of the output word length of the compression waver. ) As mentioned above, when the zero point of the transfer function of the pressure aF wave device is on the unit circle, the minimum entry is Lopy is obtained. The exact inverse transfer function is on the unit circle, pressure! Same as the zero point of the OF wave device It has Maki in the same position. Such a restored F-wave device is unstable. Such instability The accuracy is satisfactory until a bit error occurs. A bit error occurred When, the incorrect random is generated by the "initial state" K, the restored P-wave generator emits until it is saturated. Scatter. For use in configurations where the combination of compression and restoration doors is unstable. Two different systems are described below. These two systems have pit errors. In short, there are seven systems that can be used to recover from multiple actual Signal value! , is periodically transferred to the digital restoration door waver 70, and it is periodically Reset the period. Of course, this requires blocking the signal, and also Data from the point of error to the end of the block without the need for complex and high-speed logic circuits Loss is necessary.

Another system is to create a bit error signal when an error is detected. Check bits and error checking means are used. Compression using false signals The pole of the P-wave generator 70 is temporarily moved inside the unit circle. During this time, P wave device 7 0 recovers from the error and there is no need to reinitialize the F wave generator with the actual signal value f.

By placing the poles of the restored p-wave device inside the unit circle, the to-wave device becomes stable and prevents errors. The incorrect "initial state" created by Shi disappears. Under this steady state, the P-wave device is stable. a), blocking for error recovery is not necessary. A stable F of this type The wave device combination will also be explained in more detail below.

It turns out that if the zero of compression is not on the unit circle, then the exact opposite pole will be offset from the unit circle. Ru. However, as will become clear in the example below, such a combination of compression and decompression P-wave generators This is not practical for music data compression. The three mP wave device transfer function is /,00 If a point contains two real zeros near it, tX 　200　/　,! 2000 (distance along the unit circle corresponding to 200Hz *1 ) must be at a distance of: If the distance is large, low frequency components will be suppressed. Not attenuated.

Since aoo57o π; 2′″, the zero point is divided into two zeros as follows. If points are used, there must be a coefficient equal to 2-14, at the lowest end of the P-wave unit calculation. /Hiro bit is added. If no truncation is done, approximately /1 bit is double ) is added to ab, making f-t compression almost impossible.

Seven other compression-decompression P-wave device combinations are not suitable for use in this system. Also includes a configuration in which the zero point of the pressure mF wave device and the pole of the restoration f wave device both deviate from the unit circle. Ru. In these configurations, there is no truncation of the calculation word length, but the output of the pressure I; The power is 2 bits or 2 bits longer than the word length of the analog-r isotal fS device. It will be cut off immediately. If the output quantization level is equal to 90, the dojiization noise The distribution of power is Therefore, the noise becomes white. Restoration distortion is the output noise caused by the noise generator input to the restoration F-wave device. equals the sound. The variance of this output noise is given by the following equation.

Since the input noise samples are white and statistically independent, the output noise variance is can be expressed.

Here, gi is the impulse response (at the i-th sampling time) of the restored F-wave generator. ) Tilt by the lth value. In other words, the sum of squares of the in/4rus response sample The square root is the standard deviation multiplier. This multiplier can be calculated by solving an appropriate difference equation. Calculated for different pole positions.

It has become clear from calculations that this truncation of the output of the compressed ν wave generator results in If the noise power is too high or is present, the music output is concentrated in a small part of the signal bandwidth. This means that undesirable sounds are produced. Therefore, the output power of the compression door Truncation of the code is not a satisfactory method of compressing audio data.

The restored P-wave generator is decoded from the output of the compression P-wave generator 30, which is not the exact inverse of the compressed P-wave generator. There are no bit errors in the transfer up to the input of the inotal restoration waver, and the pressure is also the output of the waver 30. If no truncation is performed, the output of the Huffman decoder 6B is the output of the compression waver 30. Become the same as power. In this way, from the input of the compressing F-wave unit 30, the decompressing P-wave unit 70 is It turns out that the transfer to the output is simply the product of the transformations of the two wavers 30 and 70. Ru. Another data compression system that is an embodiment of the present invention uses a compression-decompression device. Combinations are included, and the zero point of the compression waver is to reduce the entropy. A certain point on the unit circle is placed on the unit circle, and the corresponding glue of the restored Toba device is placed on the unit circle to obtain stability. is located adjacent to the zero point inside.

The frequency characteristics and stability of such a compression-recovery door waver combination can be easily calculated. Ru. For example, a compression wave generator has two zeros at (i, o), and a restored P wave generator has two zeros at (/ Consider a compression-decompression Fm combination with two poles at −,00/riyo,0). Ru. The pole-zero notation of the cascade connection of such a compressing p-wave device and a restoring p-wave device is the first The frequency characteristics of this door transducer combination are shown in Figure 1. There is. As shown in Figure 1/, this combination reduces the cutoff frequency to / A very flat high-pass door waver of IH island is obtained.

With this F-wave device combination, recovery from pit errors is within 20 to 30 ms. be. The restored P-wave device 70 used in this specification can achieve the desired operation of the restored P-wave device. It can be configured with a digital combinator that has been degrammed to preferable.

Consideration of the word length of the restored F-wave device In order to obtain a stable restoring FM unit that operates without truncation, the calculation word length must be must be long.

For example, the above real number pole / r Hz OF waveformer has an operation of 3≠Pitt or higher ( , ooiyr = knee? )is necessary.

When the analog-digital (A/D) word length is l≠ bits, at the lowest end It is necessary to have at least one pole hit, a pig bit, and at least one pole hit/bit at the top end. With Hirogoku configuration requires a longer calculation word length. Currently, such long Computers that operate at long card lengths are impractical for civilian voice-to-data processing applications. stomach.

Fortunately, even if you truncate the decompressor's calculation word to a practical length, the system The diameter is so small that the deterioration of the system characteristics can be ignored. Analysis of truncation noise in operation words is based on quantization. This is done in much the same way as analysis. For this analysis, the noise generation with noise power q2//2 A generator (one generator per coefficient) is added to the p-wave generator input to generate the input-output 4 standard deviation The multiplier of is calculated. The value of q is the quantization level of the truncated kettle operation word.

/♂Hz 2 real-pole restoration-wave generator, the multiplier is 3227. At this time, the performance The arithmetic word length must be ~12 bits longer than the word length. Not so When the AA conversion becomes larger than the φ quantization noise power, /≠bit, 26 or 27 bits are required for the calculation of the restored P-wave generator. Become. The compression waver has (/, 0) KJ real zeros, 7 kHz on the unit circle. If you have two complex poles at the same time, set the Butterworth position at 20 Hz. 7 along the complex pole and unit circle. From the unit circle at kHz/00 Hz A restoring F-wave generator with a complex pair at the input can be used. Restoration like this Toba's standard deviation multiplier is 32, that is! It's a bit. and compression-decompression P-wave device The frequency response of the combination is the same as the first/first one except for a narrow notch at 7 kHz. The frequency characteristics are almost the same as those shown in the figure. When calculating with 24L bits, the signal There is virtually no deterioration. For example, the digital calculation length of 2≠Pitt for restoring de-waves. - Combiators can be used in the system of the invention at a reasonable cost.

Error Detection and Restoration System with Movable Pole of P-wave Unit Figures 72A and 72B A variant of the invention is shown in which a coded digitally compressed signal is recorded. Check bits are created for sound and/or transmission. Playback device and/or Or, in the receiving device, the error signal is determined by the error K detected using the check bit. is created, and using this temporarily the pole of the digital restoration ν wave device does not change the polar angle. is moved inside or further inside the unit circle in two planes. Unstable restoration F wave device In the case of This eliminates playback errors and eliminates the need to initialize the Pe device again. recovery from transmission errors and/or transmission errors. In the case of a stable restoring p-wave device, it is possible to Moving it inside the circle will accelerate recovery from false signals.

First, Figure 1.2A will be explained. Figure 12A uses a check pit. A digital recording transmission section of a variation of the data compression system is shown. 7th child ALlu's system is similar to the 1st A- system, with an analog detector. AI converter! 0, Di Zotal pressure MF wave generator 30, Huffman encoder IA0, switch 4tr, recorder 0, buffer memory! 2. Modem Yohiro and Taimin A group ♂1j control device 21A is included. These are all shown in the picture and mentioned above. It can be made into the same type as the previous one. Analog Takaoka'e, de-emphasis waver This transducer is installed on the input side of the 23K transducer and is shown in Figure 1A. It performs the same function as Deinotal door device 23.

In one form of the invention, shown in FIG. 12A, the Huffman encoded signal h(Δ, 1) is Recorder jO or modem depending on the position of switch 4t♂! Check the route connecting to ≠. A pit generator is provided. Check bit generator! Occurrence of PO A right check pit is added to the digital stream of the Huffman-encoded signal, as described above. Recorded and/or transmitted along with encoded digital compressed signals. check· A large number of pit generation and error detection using such check bits. The scheme is well known and does not require detailed explanation. Recorders and modems do not allow recording or transmission. Check bit generation for generating check bits to be added to the data stream Often includes means of production.

Recorder! The encoded digital signal and check bits as recorded at rO are Q It is played back using the playback device O shown in /2Br. See Figure 12B below. and explain. modem! 11t (12th hl, !Kl”) The number is a line! It is transmitted to modem 12 (FIG. 72B) via B. Switch 6 Hiro is again Connect the raw output or modem output to error checking circuitry 2. In this circuit 2, The word flow is checked for pit errors. When an error is detected, the pit error signal is created. Is this signal a line? Digital restoration via ≠ and switcher - sent to the wave generator, temporarily shifting the pole of the single wave generator inward.

If the error code was given by the playback device and/or the modem Remove the check bit signal from the signal. and a huff from the error checking circuit. The Mann-encoded digitally compressed signal stream h(Δ) is applied to a Hough-ry decoder 6B. . This decoder may be of the same type as described in FIG. From Huffman decoder The isotal compressed signal Δ is digitally restored via the buffer memory 72. wave generator 7 (7A).Similar to the restored F wave generator 70 in FIG. 1B, the restored single wave generator 70 A operates with truncation, converting the compressed signal input Δ1 into a sample signal of equal length /n(ou t). this! U(,,1) is the input signal of the compression single wave generator 30 (FIG. 12A). It is very close to the sample signal f. f' digital-to-analog converter 7 is reliable. Convert the number sample /n(out) to analog format /(t)out. D/A strange An analog high frequency emphasis P wave device 75A is provided on the output side of the converter.

This one-wave device is the one-wave device 7 of the 7B cause! performs the same function as In other words, this single wave device is The amplitude of the high frequency signal de-emphasized by the ν wave generator 23A is recovered.

Poles in the unit circle of two planes As mentioned above, one embodiment of the present invention uses a compression-decompression P-wave generator combination. , the zero point of the compressed one-wave generator is placed at a specific point on the unit circle in order to reduce the entropy. In order to perform stable operation, the restoring F-wave device has a point adjacent to the above zero point inside the unit circle. It has corresponding poles.

False check circuit Whenever the octopus generates a false signal, it temporarily restores the By moving the term inside the unit circle of two planes, we can correct the bit error of the restored Toba device. Recovery is accelerated.

Figure 73 shows the zeros and poles of the transfer function of the compression-decompression F-wave generator combination. Ru. The zero point of the compressed wave device is located at zero degree on the unit circle, and the pair of the restored single wave device The pole of is adjacent to the zero point and is usually located at a distance of @aooi tata within the unit circle. zero This combination of points and poles is the same as the combination of FIG. 1O described above. But the first In the system shown in Figure 13, the bit error signal is sent from the error check circuit through the line , the poles of the restoration single wave generator are temporarily moved inward. For explanation purposes, In order to rapidly recover from the loss, the pole is moved to the point aO and Otsu on the inside of the unit circle. do. After a short period of time, e.g. That is, it returns to the point aooiyr of zero degree inside the unit circle.

The difference equation of a restored single wave filter with two poles at 00 inside the unit circle is as follows: become.

7 = 2Δ. 10a7n-1=2Δ, +y! , -1-2''my,,-1)/, = yn+ a/, -1=yn+ In-, -2-”/!1-1 4214 Here a = / -2”” m; integer A restoring door barrier realizing equations (2) and (2) is shown in the fourteenth row. below , will be explained with reference to the first ≠ figure. 1 The restored door wave device 70 AK shown is /. Contains the digital multi-breather/30 of Hirotai with force/32f. . A compressed signal Δ is applied to this multi-bleph sensor/30 from a decoder 6B. Maru The output of Chibrefusa/30 is given to the arithmetic logic unit ALU/j≠.

This ALU/3 is controlled by the timing control device 76A to shift and add , the required multiplication by subtraction is performed.

The output of ALU / 3 is / to 2 digital multi f p / 31 It is connected. r multiplexer! One output of 31 is connected via line/hiro lA. Pairs of series connected system registers ≠ 0 and 7 registers /l/-2 It is connected to the. The other output of the demultiplexer is 7 fi/4 via B. is connected to the shift register/≠♂. y value determined by ALU is loaded into register 1tto, and the previous value of y is It is shifted from register /≠0 to register /lA2. 3rd register/'AI is the sample value calculated by ALU/31A/! I(out) is given It will be done.

Register 4tO, /≠2. and lIA which output is multi-breather/30t- It is applied to the input of ALU/34t via the signal. When used, stored in register/≠♂ The resulting value is 'n-t ("t).As is clear from the formula (■), the registers it, Use y and In-1(out) given from to and /≠♂, respectively. The sample value/n (o*t) is calculated.

As long as m is equal to the number of limbs smaller than infinity, the restoring F-wave generator 70 operates stably and There is no need to initialize or reconfigure the wave device. When there is no bit error, the − wave device is compared. It operates at a large value of m, e.g. m = 7, so that the pole of the single wave generator is adjacent to the unit circle It is placed at 00/Tata from the circle. Incorrect check circuit octopus Therefore, when detected, a smaller value of m, for example m=Hiro, is used.

As a result, the pole of the one-wave device moves inward from the unit circle, θot2! will be moved to the point. Mistake It is given from the check circuit 2 (12th B-) to the ALU/31A via the The bit error signal that is generated is multiplied by the coefficient J1, so simply controlling the shift amount controls the value of m used to realize equations (g) and (to). Mistake When detected, some nominal time (e.g. j Ornm ), the contents of the human LU renostar are not shifted to the right. This restores the F-wave device. The poles of are moved inward from the unit circle, accelerating recovery from the transient. This smoke After a period of time, the poles of the restored F-wave generator are again adjacent to the unit circle and normal operation is resumed.

Poles usually on a biplane focal circle As mentioned above, a seventh embodiment of the present invention uses a compression-decompression p-wave generator combination. (2) In order to reduce ci, the zero point of the compressed F wave generator is set at a specific point on the unit circle. The restoring p-wave generator operates normally, that is, when there are no bit errors. 〒, has a corresponding 〒 located at the same position as the zero point on the unit circle. However, the pit When the error is detected by the error check circuit Y2VC, the restoration Toba device 70 poles are Temporarily moved inside the unit circle, stable operation of the restored P-wave device and recovery from errors is obtained. This embodiment is based on the above-mentioned receiving device or reproducing device shown in No. 12B. This can be realized using the restoration p-wave generator 70A shown in FIG. death However, when the transient state is smooth, the restoration P-wave device 7 (7A is the unit of pole as shown in Figure 1) It operates in a circle. In Fig. 1j, the two zeros of the compressed P wave generator 3o are two squares. It is placed at zero degrees on the unit circle of the surface. When there is no pit error after setting During operation, the λ poles of the restoration p-wave generator 70h are equal to infinity at the same point 1m on the unit circle. placed at the desired point.

If there is a bit error signal from the error check circuit 9'2, the restoration F-wave generator The koshiki is temporarily moved inward from the unit circle at zero degrees. For explanation, the value of m is Suppose it is changed to ≠. Under these conditions, a restoring single-wave device quickly recovers from a mistake. However, there is no need to send the actual signal value/l to the door transducer for initial or reconfiguration. stomach.

What you have to be careful about here is that when starting the operation, temporarily change the pole of the restoration PtL device. By moving it inside the unit circle, there will be no random ramp function in its output. I'm trying to stay quiet.

Clearly, the present invention moves the poles of the reconnaissance door inward towards one position in the presence of an error signal. It is not limited to moving to. Using several values of m, from pit error The P-wave machine movement may proceed through several different landmarks during the recovery of, e.g. 2. Hiro, and using the value of 'AH equal to 7, first take the action of rn=2, then m= ≠, you can finally set m=57 and return to the initial value of m. In this case, 2 In a seventh embodiment of the present invention, the pressure-restored P A combination of wave generators is used, and in order to reduce the di/tropy, the fIF wave generator is The zero point is placed at a specific point on the unit circle of two planes, and the reconstructed P-wave device places the zero point on the unit circle. With corresponding poles in the same position, these poles are fixed and do not move inward.

As mentioned above, such a compression-decompression P-wave generator combination is unstable and transient Random output is sent out from the restoration door waver depending on the phenomenon. of such transient phenomena. In order to minimize the shadow W, by transmitting multiple actual signal values f during operation. The restored door waver is periodically reconfigured. For this type of movement, You can use the equipment shown in 1B.

The flow of signals transferred by this operation is shown in Figure 1. Below, the first I will explain with reference to Otokyo.

Switch 3! By periodically moving to the position of the broken line of t-1A, Huffman encoded difference signal h(Δ1□)...Huffman encoded signal in addition to h(Δ3+i) Signal value h(fn).

h(fQ+, n, etc. are transmitted synchronously. Switch 3! is in the position indicated by the broken line. Then, a series of actual signal values f is periodically passed through a Huffman encoder! given to AO and encoded is then blared and/or transmitted. The number of consecutive signal values f to be sent out is Equal to the order of the restored door waver.

In the signal flow denoted by 1i-, two consecutive signal values In are encoded periodically. , is used to periodically reconfigure the second-order restoring F-wave generator 70. encoded The music symbol h(7) etc. contains a label and an actual signal value ft-. The label used is The “Other” label is used to represent signals that are outside the predetermined range of compressed signal values Δ. It's different from Bell. In the first example, the label part of the encoded signal h (/) is the label , II/≠2 to distinguish it from the "other" label.

The required number of encoded signals (mh(J') are periodically transmitted as shown in the 16th row, for example, /θ milliseconds and periodically reinstalls the corresponding r digital restoration F-wave unit 70. Set. When resetting the Enomoto Toba device periodically like this, it is necessary to It is not necessary to operate the IC in a unit circle with two planes. Due to transient signals This is because the Lande function output generated by this process disappears within 0 to 10 milliseconds.

Although the present invention has been described in detail in accordance with the requirements of the T-Patent Law, those skilled in the art will understand that other types You can make various changes and transformations. For example, many of the features illustrated are This can be achieved using appropriate computer routines using a digital computer. can. It is intended to cover such modifications and variations as fall within the spirit and scope of the invention. Describe the scope of claims.

Brief description of the drawing The invention will become clearer from the following description with reference to Appendix 1. In the drawing, some Similar symbols are used to refer to like parts in the figures.

Figures 1A and 1B both show block diagrams of the data compression system. Di The digital recording-transmitter section is shown in Figure 1A, and the playback-receiver section is shown in Figure 1B. Ru.

Figure 2 shows the various locations of the data compression system shown in Figures 1A and 1B. It shows the waveform and graph of the signal that appears.

The second AIiN is provided adjacent to the input and output of the data compression system, respectively. The frequency characteristics of the high frequency DEEN7 assist F wave device and the high frequency DEEN7 assist door wave device are shown.

The third culprit is the graphical representation of the encoded difference signal; 2 shows the format used to encode the difference signal.

The first IL diagram shows the probability that a digital sample signal value occurs at a certain quantization level. This is a graph used to explain the relationship between the rate and the size of its quantization level.

FIG. 5 shows the zero points on the two-plane unit circle of the transfer function of the second-order compressed F-wave device.

The z-th transformation is a plurality of z-transforms that can be used in the compression door device that is an embodiment of the present invention. Indicates the location of the zero point.

Happy 7- is 3 with zero points at some positions shown in the 6th evil on the unit circle of 2 conversion. 3 is a graph showing frequency characteristics of different compression P-wave devices.

No.! Figure 7' shows details of the type of compressed F-wave generator that can be used in this system. It's a bad idea.

Figure 3 is a table showing types of truncated Huffman codes that can be used in the present invention. be.

Figure 1O can be used in this system and provides a stable system without error detection. The I4 turn of the zero-pole of the compression-decompression waver combination resulting in the stem is shown. are doing.

The 1st / worst is the compression-restoration F-wave device with the zero-point I4 turn shown in the 10th Shows the frequency characteristics of the combination.

Figures 72h and 12B are similar to Figures 1A and 1B-, respectively. is a check bit generator and a temporary restoration P when a bit error is detected. A system including error checking means used to move the poles of the wave generator inward It shows.

Figure 13 shows the zero-pole pattern of the compression-decompression P-wave generator combination, and shows the transient The poles of the ninth Restoration Fit device, which speeds recovery from the condition, are temporarily moved inward.

FIG. 11IL is a block diagram showing details of a restoring p-wave generator that can be used in the present invention.

Figure 1j shows seven additional pressures that can be used in a system embodying the invention. It shows the zero point-pole number of turns of the reduced-restored door waver combination.

Figure 76 shows that the actual digital signal value is periodically transferred to the restoration F-wave generator and Graph the H77 encoded signal present in the system to reinitialize the restored P-wave device. It is a precise expression.

Procedure Nefu Seishoboshiki) 3 ta January - Sunday, 1985 Mr. Michibe Uga, Commissioner of the Patent Office 3. Person who makes corrections Relationship to the incident: Patent applicant Name: Niss R.I. International 4, Agent 8. List of attached documents: 1 translation of drawings

Claims (1)

[Claims] 1. Data compression systems for processing fixed length digital sample signals In the A digital compression filter that generates a compressed signal in response to a digital sample signal. A transfer relationship that is a step and has a zero point approximately at zero degrees from the origin on the unit circle in the Z plane. has a gain that is a power of 2, and performs arithmetic operations without truncation or rounding. digital compression filtering means for carrying out; A truncated Huffman code is created in response to the output of the digital compression filtering means, and encode a compressed signal that is within a certain signal range, and encode a compressed signal that is outside the predetermined signal range. digital encoding means for labeling signals; digital decoding means; the digital decoding means for decoding the output of the digital encoding means; a means for transmitting and restoring the same in response to the output of the digital decoding means; A digital restoration filtering means for filtering is used on the unit circle or unit of the Z plane. It has a transfer function with a pole inside the circle, approximately zero degrees from the origin, and the output Digital restoration that performs arithmetic operations with truncation to generate truncation errors A data compression system comprising filtering means. 2. The transfer function of the digital compression filter means is ±41 from the origin on the unit circle of the Z plane. ．． Among angle pairs of 41°, ±60°, ±90°, ±120°, and ±180° At least/totokoro contains a zero point, The transfer function of the digital restoration filtering means is on the unit circle of the Z plane or inside the unit circle. A pole is located at an angular position approximately corresponding to the zero point of the transfer function of the digital compression filtering means. A data compression system according to claim 1, comprising: 3. Error in transferring output from digital encoding means to digital decoding means and generate an error signal when detecting an error in the transmission. , the above transfer means uses an error check code and an error detection means, In response to the error signal from the error detection means, the digital restoration filter means detects the error signal. The angle of the poles of the digital restoration filtering means is adjusted so that errors made can be easily recovered from. a means of temporarily moving the pole inside the unit circle in the Z plane without changing the 2. A data compression system according to claim 1, comprising: 4. The transfer function of the digital restoration filtering means has a pole on the unit circle of the Z plane, and the above pole is temporarily moved inside the unit circle in response to the error signal from the protection detection means. A data compression system according to claim 3. 5. The transfer function of the digital restoration filter means has a pole inside the unit circle on the Z plane, and In response to an error signal from the error detection means, the above pole temporarily moves further inside the unit circle. 4. A data compression system according to claim 3, wherein the data compression system is operated. 6. The transfer function of the digital restoration filtering means is has a pole at an angular position approximately corresponding to the zero point of the transfer function of the Tal compression filter means, and In order to periodically initialize the operation of the digital restoration filtering means, the digital restoration means for periodically applying a plurality of consecutive digital sample signals to the source filtering means; 2. A data compression system according to claim 1, wherein the data compression system comprises: 7. The digital sample signal given to the digital restoration filtering means is digital encoding means, digital decoding means, and the above-mentioned transfer means. A data compression system according to claim 6 given to the digital decompression filter means. Mu. 8. The digital encoding means is transmitted through the transfer means and the digital decoding means. A plurality of consecutive digital terminals are Claim 7 operates periodically to periodically label the sample signal. Data compression system as described in Section. 9. The above-mentioned digital encoding means is a continuous digital - Operates periodically every 6 to 16 milliseconds to provide sample signals periodically A data compression system according to claim 8. 12. The compressed signal from the digital compression filter means is connected to its sample signal input and The above evaluation digital sample signal value is related to the difference between the evaluation value and the evaluation value. Requests that can be accepted using sample signals on both sides of the digital sample signal to be evaluated. The data compression system according to item 1. 11. Digital samples by analog-to-digital conversion of analog signals analog-to-digital exchange means for signals, and In order to convert the digital output signal of the digital restoration filtering means to an analog format. digital-to-analog exchange means, 2. A data compression system according to claim 1, comprising: 12. The data according to claim 11, wherein the analog signal includes a music signal. data compression system. 13. At a speed of 30kHz to 50kHz from the above analog-to-digital conversion means. 13. A data compression system according to claim 12, wherein digital sample signals are obtained. Tem. 14. The above transfer means are means for recording the encoded signal from said digital encoding means; and means for reproducing the signal recorded by the recording means; 2. A data compression system according to claim 1, comprising: 15. The transfer means includes first and second modems, and a transmission means interconnecting the modems. A data compression system according to claim 1, further comprising a transmission link. 16. For high frequency de-emphasis of the signal input to the above digital compression filtering means. input filter, and a high frequency encoder of the signal output from the digital restoration filter means. output filter for fasis, 2. A data compression system according to claim 1, comprising: 17. Claim 1, wherein said input filter and said output filter are of digital type. 17. The data compression system according to item 16. 18. Fixed length digital data sample source, for generating compressed signals compression filtering means, in response to the regular signal from the compression filtering means, the digital filtering means; Restoration filtering means for regenerating the data signal and restoring the output from the compression filtering means. for digital data compression including means for transferring to the input of the original filtering means; system to reduce signal enlobby with little signal distortion. In the method of operating the compression tsunami means and the restoration tsunami means, (1) The transfer function is an angle of at least 0° measured from the origin on the unit circle in the Z plane. (2) the gain is a power of 2, and (3) the signal truncates operating the digital compression filtering means, and (1) The transfer function is measured from the origin on the unit circle or inside the unit circle on the Z plane. (2) has a pole at the same angular position as the zero point of the digital compression filter; The calculation word length is truncated because a truncation error occurs in the output from the digital restoration filtering means. activating the digital restoration filtering means with The operation of compression filtering means and restoration filtering means in a data compression system characterized by How to make it. 19. Located on the unit circle, measured from the origin: ±41.41°, ±60°, ±90°, at least one pair of angular positions within ±120° and ±180° operating the digital compression filter means with a transfer function having additional zeros; and Beauty On or inside the unit circle of the Z plane, same as the zero point of the digital compression filtering means. Digital reconstruction filtering with transfer function with additional poles at additional angular positions operating the means; 19. The method of claim 18, comprising: 20. Occurs when transferring the output from the compression filtering means to the input to the decompression filtering means. Accelerating the recovery of the recovery filter from the disturbance in response to a transient error. , temporarily moving the poles of the restoration filtering means inward from the unit circle in the Z plane, 19. The method of claim 18, comprising: 21. Operate the digital restoration filtering means with a pole on the unit circle of the Z plane and to periodically initialize the operation of the restoration filtering means. A plurality of consecutive digital sample signals are periodically applied to the restoration filtering means. The number of consecutive digital sample signals obtained depends on the order of the restoration filtering means. to be equal, 19. The method of claim 18, comprising: 22. Data compression systems for processing digital sample signal streams In the Digital compression that generates a compressed signal in response to a digital sample signal A tsunami means whose transfer function is 0° as measured from the origin on the unit circle of the Z plane. and ±41.41°, ±60°, ±90°, ±120° and A digit that has a zero point in at least one pair of angles of ±180° compression filter means; Almost the same as the zero point of the digital compression filtering means on or inside the unit circle on the Z plane. idigital reconstruction tsunami means with a transfer function with poles at the same angular position, and for transferring the output from the digital compression filtering means to the digital decompression filtering means; means, A data compression system comprising: 23. The above transfer means is a truncated Huffman code of the output from the digital compression filter means. Huffman encoding means for encoding, and and means for decoding the output of the encoding means in response to the output of the encoding means. 23. The data compression system according to item 22. 24. The transfer means is a bit check generation means and an error occurs when a transient error exists. means for generating a signal, and Under the control of the above error signal, recovery is performed to rapidly recover from the detected transient error. A method of temporarily moving the pole of the original filtering means inside the unit circle without changing its angular position Step, 23. The data compression system according to claim 22, comprising: 25. Data compression systems for processing digital sample signal streams In the Digital compression responsive to a digital sample signal to generate a compressed signal A digital filter whose transfer function has a zero point on the unit circle of the Z plane. compression filtering means; If the poles of the transfer function are on or inside the unit circle on the Z plane, digital compression is not possible. digital restoration filtering means at approximately the same angular position as the zero point of the filtering means; means for transmitting the output from the compression filtering means to the digital restoration filtering means; and A filter for transmitting the output from the digital compression filter means to the digital compression filter means. In response to a transient error, the poles of the restoration filtering means are moved to the Z-plane without substantially changing their angular position. Easier recovery from transient errors by temporarily moving the surface inside the unit circle means, A data compression system comprising: