JPS61158220A - Signal transmitter - Google Patents

Abstract

PURPOSE:To prevent the transmission of errors and to avoid the generation of noises, etc. by clipping a requantized bit with the positive or negative maximum value when this bit has an overflow and at the same time feeding back an error produced then the perform a noise shaping process. CONSTITUTION:The differential filters are formed with estimating units 12A-13D and adders 13A-13D to obtain the estimated errors of four systems for an encoder of an audio bit rate reduction system. An estimating range adaptive circuit 21 selects the most suited one of those differential filters. The differential output, i.e., an estimated error is sent to a bit compressing means consisting of a shifter 15 of gain G and a quantizer 16 via an adder 14. Then the quantizer 16 performs the requantization to extract several bits out of the prescribed positions of data. The quantized output is sent to an adder 18 of a noise shaping circuit 17 via a clipping circuit. A quantized error thus obtained is sent to and estimating unit 20 via a shifter 19 for feedback of the error.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a signal transmission device that transmits a signal obtained by digitizing an analog signal such as a PCM signal by dividing it into blocks of a certain number of words, and in particular, The present invention relates to a signal transmission device that performs conbanding processing to reduce the transmission bit rate.

[Conventional technology]

In recent years, analog audio and video signals are often sampled, quantized and encoded, and then transmitted, recorded, and played back as so-called PCM (pulse code modulation) signals. ing.

When transmitting, recording, and reproducing such PCM signals, etc.1, in order to obtain a band of, for example, about 20 kHz and an S/N of about 90 dB or more, the sampling frequency fs is set to 44.1 kHz, and linear quantization of 16 bits per word is performed. is generally adopted, but the transmission rate in this case is 7
It is extremely high, reaching over 00KBPS (700 bits per second).

By the way, digital signals obtained by A/D conversion of analog signals such as audio signals and video signals as mentioned above have biased statistical properties and parts that are less important from the viewpoint of audiovisual phenomena. It is possible to compress the amount of information by using Arakokichi, for example, difference/integration processing, compression/expansion processing (conbanding processing)
It is known that there is very little deterioration in signal quality even if this is done.

Taking these points into consideration, the applicant first created blocks for digital PCM signals, for example, by a certain time unit or by a certain number of words, and performed predictive processing such as differential processing and combining for each block. A patent application filed in 1983-1997 was issued to transmit, record, and reproduce the data through processing.
No. 687-9, Japanese Patent Application No. 163054, 1973, Japanese Patent Application No. 1973
No. 8-166267 or patent application 1982-210382
It is proposed in the No.

In these techniques, at least one
Word standard data, such as straight PCM arcs, is provided, and by performing arithmetic processing such as sequentially adding difference data based on this standard data, the original sampling data (stray PCM data) in the block can be extracted. All can be restored. Furthermore, as the above-mentioned conbanding process, it has been proposed to perform noise shaping processing by requantizing the input data and feeding back the predicted value of the quantization error at this time (so-called error feedback). It is desirable to perform this quantization error prediction process separately from the signal prediction process in order not to degrade the instantaneous S/N ratio. In this case, the extraction position of the requantized bits with respect to the original data bits during the requantization, the so-called ranging position, is determined based on the data before the noise shaping process.

[Problem that the invention seeks to solve]

By the way, in order to further increase the bit rate reduction efficiency, when transmitting or recording/playing without providing the reference data for each block, when the input signal level suddenly changes near the block boundary, the noise - Due to shaping, the error fed back from the last word of the previous block is superimposed on the first word of the next block, potentially causing an overflow in the requantized data. This overflow causes adverse effects such as distortion on the transmitted signal.

In view of these circumstances, the present invention has been proposed to divide an input signal into blocks and transmit them without providing a reference word for each block, and to perform signal and noise prediction processing separately. An object of the present invention is to provide a signal transmission device that can prevent the adverse effects of overflow of requantization bits caused by feedback of errors from the previous block when the signal level changes near a block boundary. do.

[Means for solving problems]

That is, the signal transmission device of the present invention includes means for dividing an input digital signal into blocks of a certain number of words along the time axis, performing predictive processing on the signal for each block, and reproducing the predictively processed signal. means for performing noise shaping processing by quantizing and feeding back the quantization error; and ranging during the requantization based on the maximum absolute value within the block of the predictively processed signal. A means for determining the position, that is, a position for extracting bits of requantized data with respect to all bits of data before requantization, and a means for determining the positive and negative values of the data before requantization when an overflow occurs in the requantized data. The above problem is solved by including means for clipping the requantized data to a positive or negative maximum value depending on the polarity.

[Effect]

By attaching requantized data to the maximum positive or negative value when an overflow occurs, the next feedback
The amount of errors is reduced, and error propagation can be reduced.

〔Example〕

Schematic Configuration First, the overall schematic configuration of an audio bitrate reduction system, which is an example of a signal transmission device to which the present invention is applied, will be described with reference to FIG.

The system shown in FIG. 1 consists of an encoder 10 on the transmitting side (or recording side) and an encoder 3 on the receiving side (or reproducing side).
0, and the input terminal 11 of the encoder 10 is an audio PC obtained by sampling an analog audio signal at a frequency fs, and performing quantization and encoding.
M signal X (river is fed. This input signal x(n)
are sent to the predictor 12 and the adder 13, respectively, and the prediction signal x from the predictor 12 (nl is the adder 1
3, by subtracting the prediction signal - (river) from the input signal x (nl, the prediction error signal or (broad sense) difference output d(nl, that is, d(river =
x(nl-x(river)
...■ is output.

Here, the predictor 12 generally uses p past inputs X(n
-p ) l X(n-p+1) l ・I X(n
-1) to calculate the predicted value x (nl, where αk (k=1, 2..., p) is a coefficient. Therefore, the above prediction error output ) difference output d (nl is d(n)=x(river-Σα-x(n-k)...
・・・・・・・・・■ can be expressed as -1.

In addition, in the present invention, data within a certain period of time of the input digital signal, that is, the input data is divided into blocks for each certain number of words g, and the coefficient α is adjusted so that the optimum predictive filter characteristics can be obtained for each block. The set of is selected. As will be described later, this can be considered as having multiple filters for obtaining differential outputs (prediction error outputs), including predictors with different characteristics or adders, and these multiple differentials The optimal filter among the processing filters is selected for each block. The selection of this optimal filter is based on the maximum absolute value (peak value) of each of the plurality of differential processing filters or the maximum absolute value (peak value) of the
Or the value obtained by multiplying the maximum absolute value (peak value) by the coefficient,
This is done by comparing each other in the prediction/range adaptation circuit 21, and specifically, the difference processing filter that minimizes the maximum absolute value (or its coefficient multiplication value) is the optimal one for the block in question. Selected as a filter. The optimum filter selection information at this time is output as mode selection information from the prediction/range adaptation 1λ circuit 21 and sent to the predictor.

Next, the difference output d as the prediction error is sent via an adder 14 to a bit compression means consisting of a shifter 15 with a gain G and a quantizer 16, and is expressed as a floating point, for example. Compression processing or ranging processing is performed such that the exponent part in the form corresponds to the gain G and the mantissa part corresponds to the output from the quantizer 16. That is, the shifter 15 converts the digital binary data into the gain G. The so-called range is switched by shifting (arithmetic shift) by the number of bits corresponding to The requantized output is taken out via a clipping circuit (clipper) 26, which is the gist of the present invention, and then sent to a noise shaping circuit (noise shaper) 17.
is sent to the adder 18. Noise shaping circuit (
The noise shaper) 17 uses an adder 18 to obtain a so-called quantization error (and overflow error) corresponding to the error between the output and input of the quantizer 16 (and clipper 26), and uses this quantization error as a gain G. Predictor 2 via shifter 19 of
0, and a so-called error feedback is performed in which the predicted signal of the quantization error is fed back to the adder 14 as a subtraction signal. Next, the prediction/range adaptation circuit 21 outputs range information based on the maximum absolute value within the block of the differential output from the filter of the selected mode, and sends this range information to each shifter 15 and 19 for each block. The above gains G and G are determined as follows. Also, the predictor 20
The characteristics are determined by the transmission of the mode information from the prediction/range adaptation circuit 21.

Here, as a noise prediction process after the adder 14, the basic operation when no overflow occurs will be explained.

The output d from the adder 14 (the difference output d (b) is
The predicted signal e of the quantization error from the noise shaper 17 (d after subtracting the river d (n) = d (nl-e (nl
・・・・・・・・・・・・・・・■The output d (nl) from the shifter with gain G
-d (nl ・・・・・・・・・・・・
...■. In addition, the output d(kawa) from the quantizer 16 (via the clipper 26) is expressed as ...
・・・・・・・・・■ becomes Noise Shaper 17
In the adder 18 of The predicted signal e (n) is as follows. This equation 0 has the same form as the above equation 0, and the predictors 12 and 20 each have a system function of F'IR (finite impulse response) of It's a filter.

From these 0 to 0 equations, the output △ d (nl) from the quantizer 16 is a(n) −G ・(d(n) − e(nl) + e(nl
By substituting the above equation 0 into d(river) of this equation 0, we get the output d(nl) which is taken out via the output terminal 22.Here, the above x(nl, e(nl
, d(nl) respectively as X(zl, E(
z), and the sum (・), then e(z) = G, X(Z
l(1-day., 2-k)k++l.

The range information from the prediction/range adaptation circuit 21 is taken out from the output terminal 23, and the mode selection information is taken out from the output terminal 24.

Next, the output '>(nl) from the output terminal 22 of the encoder 10 is transmitted to the input terminal 31 of the decoder 30 on the reception side or reproduction side, or the A signal Aa'+,) is supplied. This input signal d( is sent to an adder 33 via a shifter 32 with a gain of G-. The output from the adder 33 is X(
The river is sent to the predictor 34 and becomes the predicted signal x(nl,
This predicted signal x (n) is sent to an adder 33 and outputs Q, , n from the shifter 32.

It is added to Δ′. This addition output is outputted from the output terminal 35 as the decoded output x(nl).

Further, the range information and mode selection information outputted from the respective output terminals 33 and 24 of the encoder 10 and transmitted or recorded/reproduced are inputted to the respective input terminals 36 and 37 of the decoder 30, respectively. The range information from the input terminal 36 is sent to the shifter 32 to determine the gain G, and the mode selection information from the input terminal 37 is sent to the predictor 34 to determine the prediction characteristics. The prediction characteristic of this predictor 34 is selected to be equal to the characteristic of the predictor 12 of the encoder 10.

In the decoder 30 having such a configuration, the output d (nl is Q7n) from the shifter 32 is expressed as follows:
・・・・・・・・・＠／＼′, and the output of the adder 33 is
・・・・・・・・・・・・・・・It will be bad. Here, since the predictor 34 is selected to have the same characteristics as the predictor 12 of the encoder 1°, σ or Δ′ is obtained from the equation 0. Next, if the two transformations of x (nl, d (nl are respectively X(z
)・z−11(=1 Therefore, it becomes .Here, assuming that there is no error in the transmission path or recording medium and let G'(zl - public(z)), from the above equation (b) and equation 69, it becomes .

From this equation 0, it is clear that a noise reduction effect of G can be obtained for the quantization error E(z), and at this time, the spectral distribution of noise appearing in the decoder output is expressed as N(z
If it is l, then it becomes.

Further, in such a system, the G acts to normalize with a value related to the maximum absolute value within the block, but this G has frequency characteristics. Here, to simplify the explanation, the above G is expressed as G=Gp-Gf...
...Represented as the product of two elements Gp and Gf such as 0.

Of these two elements, Gp means the predicted gain due to the above predictive filter processing, that is, the amount of improvement in instantaneous S/N, and Gf means the amount of gain control due to the above ranging process, that is, the amount of expansion of the dynamic range. do. therefore.

Gp depends on the input signal frequency and not on the input signal level, whereas Gf does not depend on the input signal frequency but on the input signal level. Also Gp
has an S/N improvement amount of・The spectrum looks like the above equation 0. Regarding G(, it corresponds to quasi-instantaneous companding in which normalization is performed using the maximum absolute value within the block in the mode selected for each block.

Specific Configuration Example Next, FIG. 2 shows a more specific configuration example of the encoder 10 of the audio bitrate reduction system shown in the first part, and the parts corresponding to those in FIG. have the same reference numbers.

In FIG. 2, the predictors 12 include a plurality of predictors, for example, four predictors 12A, 12B, 12C, and 12D.
is provided. The prediction outputs from these predictors 12A to 12D are sent to adders 13A to 13A as subtraction signals, respectively.
13D and subtracted from the original input signal. That is, the four predictors 12A to 12D and the adders 138 to 13D each constitute four differential processing filters for obtaining four systems of prediction errors. Here, each of the predictors 12A to 12D apparently has a quadratic configuration, with coefficients α1. As α2, the prediction er 12
A 1. 2, 12 B Kg, 2, 12
However, by setting at least one of the coefficients of the desired predictor to 0, prediction characteristics of first order or lower can be obtained.

Therefore, for the four differential processing filters as well,
Although they are apparently configured to take second-order differences, it is possible to obtain characteristics (including those that output straight PCM data) that take first-order or lower differences for desired difference processing filters.

The outputs from these valley difference processing filters, that is, the outputs from each adder 13A to 13D, are sent to e-word delay circuits 41A to 41D and maximum absolute value hold circuit 4, respectively.
2A to 42D, and the g word delay circuit 41A
Each output from ~41D is a mode changeover switch circuit 43.
are sent to each of the selected terminals a to d. That is, since one block is an e word, a delay of one block is performed in the e word delay circuits 41A to 41D, and while this delay is being performed, each maximum absolute value (peak) hold circuit 42A The maximum absolute value within the block is detected at ~42D. These intra-block maximum absolute values are sent to the prediction/range adaptation circuit 21 and compared with each other, and the one with the minimum value is selected. At this time, after multiplying the maximum absolute value within each block by a predetermined coefficient and performing so-called weighting, the maximum absolute value within the minimum block is determined from among the data for each block from the separation filter. Mode selection information for selecting one block of data obtained is output, and this mode selection information is sent to the change-over switch circuit 43, thereby transmitting the selected one block of data to the delay circuit that outputs the data. A switched connection is made. The output from the changeover switch circuit 43 is sent to the adder 14.

The mode selection information from the prediction/range adaptation circuit 21 is also sent to the predictor 20 and the output terminal 24. Here, the predictor 20 is selected from among the predictors 12A to 12D with characteristics equal to the one selected above, in order to whiten the noise appearing in the decoder output (see the [phase] formula above), for example. Ru. That is, the predictor 20
also has an apparently second-order predictor configuration, and the coefficient β0.
The coefficients Ka and Kb corresponding to β2 are calculated by the predictors 12A to 12A.
For each set of 12D coefficients Kl, one of 2 to Kq, Ks that is equal to the -1':j- coefficient of the predictor of the differential processing filter specified by the mode selection is selected.

Furthermore, in the specific example of the decoder 30 shown in FIG. 3, the predictor 34 has an apparently quadratic configuration corresponding to the predictors 12A to 12D in FIG. 2, and each coefficient Kc,
As Kd, the set of coefficients Kl, Kz~ of the predictors 128~12D? , and 8 are selected according to mode selection information from the input terminal 37.

The other configurations in FIGS. 2 and 3 are the same as those in FIG. 1 described above, and therefore their explanation will be omitted.

Note that the encoder 10 having the above-described specific configuration
As for the hardware configuration of the decoder 30, for example, it is not necessary to actually provide a plurality of predictors 12A to 12D, etc., and the coefficients of one predictor may be switched and used in a time-sharing manner. Of course, the entire decoder 10 and decoder 30 can be realized in software using a system including a DSP (digital signal processor), memory, and the like.

2l-1t First Embodiment Here, as a first embodiment of the present invention, the sampling frequency fS of the audio PCM signal supplied to the input terminal 11 is 1.8.9 kHz, and 1 word is 16 bits. The number of words l in the block is 28 words, and the quantizer 1
6, the case of requantization into data of 4 bits per word will be explained in detail. At this time, the sets of coefficients Kz, K2 to Kq of the secondary predictors 12A to 12D are
, Ks,! : For example, i 2A : Ki=1.8426. K2=0.
86491'l B: Ka=0.875, K
4 =01'l C: K5=1.5155. K6=
0.8112D: Ky=Q, and %=O are set in advance. At this time, the transfer function 1-P(z) of the differential processing filter in each of the above modes is A: 1
-1.8426z +0.8649z-B: 1-
0.875 z-” C: 1-1.5155 z-1+〇, 81 z-2D
= 1, and the frequency characteristics of each of these differential processing filters are
The result will be as shown in multi-tracks 1A-D in Fig. 4.

That is, the difference processing filter (consisting of the predictor 12A and the adder 138) corresponding to the characteristic curve A is the second-order difference PC
This is a filter corresponding to M mode, and the amount of improvement in low-frequency prediction gain, that is, instantaneous S/N is large. A differential processing filter corresponding to characteristic curve B (predictor 12B and adder 1
3B) is a filter corresponding to the first-order difference PCIVI mode, and the difference processing filter (predictor 12C and adder 13C) corresponding to the characteristic curve C has a large prediction gain in the middle range. Predictor 12D and adder 13
Since the coefficients 7°K8 and D are both 0, the differential processing filter consisting of D has no frequency characteristics and has a so-called flat pass characteristic with a reference gain of 1, as shown in the characteristic curve in Figure 4. This corresponds to straight PCM mode.

FIG. 5 shows a specific example of the word structure transmitted for each block, including requantized audio data words WO to W27 of 28 words (4 bits per word) and 4 bits (16 bits per word). A so-called range information word WR indicating the extraction position (ranging position) of 4 bits during block-by-block requantization, and a mode selection indicating which of the four modes corresponding to the above four filters has been selected. The information word WM is transmitted in monthly blocks. Therefore, the transmitted audio data 1
The average number of bits per word is (4 x 28 + 4 + 2) ÷
28==4,214 [bits].

In Fig. 4, when a single sine wave signal is input, the input signal frequency from O to f is the characteristic curve A.
A filter having a characteristic curve C is selected for frequencies from fl to f2, and a filter having a characteristic curve C is selected for frequencies from f2 to approximately fs/2. For input signal frequencies fS/2 or higher, an LPF is applied before A/D conversion to prevent so-called aliasing.
Of course, it is removed in advance by (low bath = 23-ilk).

The frequency response of each filter selected in this way becomes the predicted gain at that frequency, that is, the amount of improvement in instantaneous S/N, and the amount of improvement in instantaneous S/N with respect to frequency is as shown in the shaded area in FIG.

However, since the actual audio input signal is a signal with a complex spectrum, selection based on clear boundaries such as "double" is not performed, and filters with characteristic curve B are also used relatively frequently.

In addition, the filter selection described above is performed when the maximum absolute value (peak value) within the block from the filter of each mode is directly compared, but the By multiplying and comparing the coefficients of
More filters for outputting CM data may be selected. In this case, as an example of the above coefficient, the peak value of the second-order differential PCM data from the filter with characteristic A is multiplied by a coefficient of 2, and the peak value of straight PCM data from the filter with characteristic A is multiplied by a coefficient of 0°7. The peak value of the data from other filters is multiplied by a coefficient 1 (or without being multiplied by the coefficient), the values are compared with each other, and the filter whose value is the minimum is selected. Multiplication of such a coefficient corresponds to translating the corresponding characteristic curve in the graph of FIG. 4 in the vertical axis direction (changing the response value). For example, the coefficient Multiplying by 2 transforms the characteristic curve A into the fourth
This corresponds to an upward translation of approximately 6 dB in the figure.

Therefore, as a result, the filter switching frequency f1 and f
2 will shift to the lower frequency side, and filters with characteristic C will be selected more frequently than filters with characteristic C, and filters with characteristic C will be selected more frequently than filters with characteristic C. .

Note that on the decoder side, the input audio data word Wo-W2? is subjected to decompression processing in block units by the shifter 32 or ranging processing, which is opposite to that on the encoder side, based on the data in the range information word WR, and then in a filter consisting of an adder 33 and a predictor 34, the mode selection information is Reverse side processing is performed on a block-by-block basis based on the data of word WM, and the original straight PCM data is restored.

Furthermore, in noise prediction processing, feedback
Due to the error, the quantizer 16 is attached to the maximum positive or negative value (that is, the maximum value that can be expressed with 4 hits of requantized data) according to the overflow, and by performing ripping processing, the quantization error (error) is removed. It's kept small. In addition, at the same time as this clipping processing, processing to limit the movement of the ranging position (re-quantization bit extraction position) toward the LSB side, as described later, or when determining the ranging position, the maximum absolute value within the block may be It is preferable that at least one process is performed, such as multiplying γ by a coefficient γ of 1 or more in advance, and determining the ranging position according to the multiplication result. These processes will be explained in detail in the second embodiment.

According to the system of this first embodiment, it is possible to transmit sound quality of low to medium fidelity, and it is possible to transmit ordinary audio signals at an extremely low bit rate (4.214 bits per word, per channel). Transmission bit rate approximately 79.6k
bits per second).

Second Embodiment Next, as an example of a system capable of transmitting music signals with medium to high fidelity (middle-high fidelity) sound quality, predictions corresponding to characteristics iA, B, and D in FIG. Using three types of filters having coefficients, the sampling frequency fs is set to the first
A case will be explained in which the frequency is set to 37°8 kHz, which is twice as high as that in the embodiment. Other specific numerical values and specific configurations are the same as those in the first embodiment.

In this case, the frequency characteristics of the three types of filters become as shown by characteristic curves A, B, and D in FIG. 6 as the sampling frequency fs is doubled. That is, the characteristic curve B corresponds to the straight PCM mode, the characteristic curve B corresponds to the first-order difference mode, and the characteristic curve A corresponds to the second-order difference mode.

By the way, when requantizing one word from 16 bits to 4 bits in the quantizer 16, 4 bits at a predetermined position are taken out.
Since the range (bit shift amount) is determined based on the data before noise shaping processing, that is, the peak value within the block of the selected mode, the noise shaper 17
As a result of the feedback error being superimposed in the adder 14, there is a risk of data overflow, especially when data is input close to the maximum value of the bits extracted during ranging processing.

To prevent this, a predetermined coefficient γ of 1 or more is applied to the peak value (maximum absolute value) in the block of the selected mode.
is multiplied in advance, and this multiplication result is regarded as the peak value to determine the ranging position, that is, the extraction position of 4 bits out of 16 bits. In this way, since the ranging position is determined by the value multiplied by the predetermined number γ of the true peak value,
Even if the error from the noise shaper 17 is fed back, the above-mentioned overflow is less likely to occur. In this case, the coefficient γ is preferably set according to the predictor characteristics of the selected mode.

Here, the predictor 20 in the noise shaper 17 has characteristics equal to those of the predictors 128 to 12D for differential processing, and is selected according to the mode selection.
In the M mode, the coefficients Ka and Kb in FIG. 2 are both 0, the error feedback amount is 0, there is no influence of noise shaping, and the coefficient may be set to γ-1. In addition, in the first-order difference mode, considering the fact that the noise spectrum at the encoder output after noise shaping processing is equal to the characteristic curve B in Fig. 6,
．． In the second-order differential mode, considering the characteristic curve A in FIG. 6, γ may be set to 1°33.

That is, when the first-order difference mode is selected, the noise spectrum at the encoder output is approximately as shown in FIG. This is the instantaneous S/ when requantizing with 4 bits.
Since N is approximately 24 dB, when the full scale (maximum level that can be expressed with 4 bits) is set to 0d13, the noise level before noise shaping is -2.
4 dB, which is subjected to noise shaping processing using the error footbank of the first-order measurement, resulting in a noise level with a spectral distribution as shown in the shaded area in FIG. Therefore, the noise around the frequency fS/2
The level is about 6 compared to the level before noise shaping.
The signal is lifted by dB, which causes the above-mentioned overflow.

This is due to the fact that the amplitude distribution of the quantization noise before noise shaping is within the 100 hits of L8B of the above requantized 4 hits, so the noise when taking the first difference is The maximum amplitude is +1LSB-(-factor LS
B)=+ILSB gives ±ILSB, which corresponds to about 6 dB rise around fs/2. Therefore, even when unmanned, +6 dB of noise is transmitted, and f
This means that the noise peak value may exist at a position of about -18 dB near s/2. Here, -18dB is approximately 0.1
25, by suppressing the peak value of the signal to 1-06125, that is, 0.875 times, it is possible to prevent overflow due to noise shaping. Therefore, the multiplication coefficient γ for the peak value for determining the ranging position may be 110.875=−1,14.

Next, regarding the multiplication coefficient γ for the peak value for ranging position determination when the second-order difference mode is selected, the rise in the characteristic curve A of FIG. 6 near fs/2 is approximately +12 dB.
Therefore, the full scale of 4-bit requantization O
dB, the peak value of the noise level is approximately -12 dB
It can exist in the position of -1,2dB is approximately 0.
25, so the peak value of the signal is 1-0.25=0.
By suppressing it to 75 times, overflow due to noise shaping can be prevented, and the above coefficient γ is 1.10.
．． 75, it becomes about 1.33.

By the way, if the ranging position during requantization changes rapidly due to a sudden change in the signal, overflow may occur due to the noise of the previous block being carried over to the next block. . This is particularly likely to occur when the signal level drops rapidly near the block boundary, and in this case, the ranging position, that is, the 4-bit extraction position for requantization is
.

As shown in B, there is a sudden shift from the MSB side to the LSB side of the 16 bits, but the data of the previous block (Fig. 8A)
Since the error amount is fed back to the data of the next block (FIG. 8B), the four-hit data extracted by requantization overflows. This corresponds to the fact that the gain G of the shifter 15, which was small in the previous block, increases rapidly in the next block.

Therefore, in the present invention, overflow is allowed, and when an overflow occurs, the code is fixed to the maximum positive or negative value that can be expressed with 4 bits of requantization, and so-called clipping is performed. This prevents reversal and minimizes the generation of abnormal noise.

Prior to this clipping process, a restriction is placed on the movement of the bit extraction position when the range becomes smaller (the gain G becomes larger). For example, as shown in FIG. It is desirable to use the position moved to the side as the bit extraction position or ranging positionG).

In this way, by limiting the amount of movement when the ranging position moves from the MSB side to the LSB side of the original 16-bit data, sudden range changes can be prevented, and overflow due to noise shaping can be prevented. The size of can be reduced.

Here, a description will be given of the effect of limiting the ranging position movement when the above-mentioned second-order difference mode is selected, in which overflow is most likely to occur, that is, the steepest noise shaping process is performed.

The noise spectral distribution at the encoder output when this second-order difference mode is selected appears in a curve shape similar to characteristic curve A in Figure 6, and the noise peak value around fs/2 is higher than that before noise shaping. is lifted by approximately 4 times, or approximately +12 dB. Therefore, when the full scale of the 4-bit requantized data is OdB, the peak value of noise can exist at a position of -1 or 2 dB, and noise occupies 0.25 dB.

Next, regarding the movement of the ranging position, as described above (
B: Since there is a restriction that the LSB side must be within 1 bit per block, the magnitude of the noise fed back in the next block is OdB from the full scale at this time.
As a result, the maximum value is 16 dB of the above 1 bit, or 0.5. In addition, if the above limit is not set, the maximum range change step is 12 bits, so -
12+6X12=60, or about 60 dB of noise will be propagated. On the other hand, if the above limit is set, the noise will occupy at most 0.5 in the transient part, and the multiplication coefficient γ is not set for the signal (
Even when γ=1), the sum of the maximum signal value 1 and the maximum noise value 0.5 is 1.5, that is, an overflow of about +3.5 dB is sufficient.

As mentioned above, this kind of ranging restriction is performed when the signal level suddenly decreases, and at this time, the S/N deterioration due to the range not being able to decrease suddenly occurs immediately before that. Due to the so-called temporal masking effect, which is masked by high-level signals, there is almost no problem with the auditory sense.

Next, after performing such ranging restrictions, the ripping process as described above is performed.

Here, clipping processing means that when the above four overflows occur, the data is fixed or clipped to the maximum positive or negative value of the 4 bits extracted by requantization, thereby preventing sign reversal and preventing errors. However, by feeding back (error feedback) the error that occurs at this time, that is, the difference between the true value and the clipped pressure or the negative maximum value, and carrying it over to the next time, that is, the normal It has been confirmed that by performing error feedback as is, there are almost no adverse effects such as distortion due to clipping.

On the other hand, if error feedback is stopped during the clipping described above, noise shaping is temporarily not applied, and the decoder output contains a large level of low-frequency noise, that is, noise according to the characteristics of the decoding filter. occurs and the resulting distortion is propagated, causing a serious problem in terms of hearing.

The difference in operation depending on the presence or absence of error feedback during the ripping process will be explained below.

Here, in order to simplify the explanation, a
Consider the following noise shaping case.

In FIG. 9, the output d(nl) from the shifter 15 is requantized from 16 bits to 4 bits in the quantizer 16, and becomes the output d(river), which controls the clipping circuit (clipper) 26 at the time of overflow. The output d (shall be a river) through the quantizer 1
The input d(river) to the output from the clipper 26 is input to The signal is supplied to the adder 14 as a subtraction signal.

In contrast to the encoder having the configuration shown in FIG. 9, the main part of the decoder is configured as shown in FIG. 10, and the predictor 34 is composed of a one-word delay means 28.

Now, as time passes, when moving from an arbitrary block (first block) to the next block (second block), the peak value of the signal becomes smaller, and the gain G of the shifter 15 changes from G to G-. 2 (but> 1), the final encoder output data of the first block is '9(nl-(d(n)-e (n-1) ・G-'')
−G0e(river=d(river−G−e (n−1)10e(river) When 0, the output plan from the chic 32 of the decoder “(nl is Δ″ d (nl= d(river+(e(nl-e(n-1,)l
-o ------@ becomes. Next, the encoder output take (n+1) at the beginning of the second block is as follows: △ d (nt-1) = (d (n+1) - e(kawa・o
)-o-y tene(n+1) =d(n+1)-G-! ? -e(kawa・y1e(n+1)
・・・・・・・・・・・・・・・(c) Also, the take/%fl, soil, ) in the decoder is △〃 d(n+1)−dCn+1)+e(n+1)G −′
i −e(nl・G・・・・・・・・・・・・・・・C
and Δ′ Δ′. Here, the decoder output x(nl, x(n+
1) When considering a・-1)=Y(・-1,)+e(n-]doorG-・, from equation 0, Qnl=Y(n-1)+d(n soup e( River G-1...
・・・・・・・・・・・・ [Phase] Also, from this [Phase] formula and 0 formula, ';" (, n+1. ) = y (n 1 ) + d (
n juice d (n+1) 10e (n+1)・G-p
・・・・・・・・・・・・・・・[phase]. As shown in these (c) [phase] equations, when there is no overflow, no interference occurs between blocks, and a large quantization error in the previous block will not trail to subsequent blocks.

Next, when an overflow occurs in the first word of the second block, e (nl
・The term G causes overflow, but
If we rewrite equation (c) by setting E to be the error amount when overflow occurs and causes clipping, we get G(n+1)-J d(n+1)-e(n)-G
-') ・G-90e (n+1)10E =d (n+1) ・G-96(River-fl+e (n
+1) 10B・・・・・・・・・・・・・・・O Therefore, the above expression [exist] is Q7. , n+1 )−d (n+1 )+(e(n+
1') 10E)-G-"-y-"-e(Kano G
・・・・・・・・・・・・・・・[phase]. Also, the decoder output of the above [phase] equation is Δ' x (n+1)=Y (n-1)+d(n)+-d
('+1) 10 (e (n+1) 10E)・G - G
......[Phase], and the difference between this φΦ formula and the above [Phase] formula is E-G -7...
...[Phase] This is the distortion that appears in the decoder output due to clipping.

Next, regarding error propagation by clipping, if the error occurring in the above formula 0 is fed back normally (
The encoded output corresponding to n+2 is '9(n+2)-[d(n+2)-(e(n+1)+E
l・G-”-9”]・G-2+e(-n-4-2)
・・・・・・・・・・・・・・・ [phase] Yo-de, C
l2) is d'(n+2)=d(n+2)+e(n+2)-G-1
-p-”(e(n+1)×E)”04 Song・-@From this equation 0 and equation (c), the decoder output is 'Q'(n+
2)-y(n 1) publication (nl-1-d (11+1)
Issue (n+2) 10e (n+2)・G −7・・・・・Music・[phase], and the influence of E disappears. On the contrary,
If the error that occurred in the above formula 0 is not fed back, △ d (n + 2) = d (n + 2) G4 + e (n + 2) ・
−−−−−■'67n+2)−d(n+2)+e(n+
2) G-"・p-'...-@This [phase] formula and the above [phase]
From the formula, the decoder output is ``'(n+2)=y(n
-1)+d(nH-d (n+1) published (n+2) 10e (n+2 houses G -9+(e (n+1.) 10E
1・G-”・y-゛ ・・・・・・・・・
[phase], and as a result, the influence of not returning E+e(n+1) will propagate.

As is clear from the results of the above considerations, it is clear that it is better to feedbank errors due to clipping as well as normal quantization errors.

In this case, only the word that caused the overflow generates the distortion E-G-jp in the above [phase] equation, and the error does not propagate to subsequent words.

Next, the effect of limiting the amount of movement when the ranging position, that is, the requantization bit extraction position moves toward the LSB side, will be explained.

First, the above equation 0 shows that the error due to clipping is fed back to the next word. This 0
The error returned in the formula is (e(n+1) + E
l·G−2 terms, which can cause an overflow in n+2 words. That is, as shown in FIG. 8B, the requantization bit extraction position (
When the ranging position) moves to the LSB side, dCn+2)<<(eCn+1)+El・G−1−p−
”, and the error E in word n+1 cannot be absorbed in word n+2, causing an overflow in the opposite direction. In this way, errors due to overflow occur one after another. This propagates, resulting in a large distortion. This situation is shown in Fig. 11. In Fig. 11, the first word in the block when the ranging position suddenly moves to the LSB side is taken as WO,
The quantization error (error) for the previous word (the last word of the previous block) is superimposed on the data DO of this word WO, resulting in the true value Po. Because it overflows beyond the range of full scale FS that can be expressed by quantized bits,
The above clipping process is performed, and the output value QO is pasted to the maximum positive value. The quantization error go at this time is
Output value Q〇−true value Po. This error Eo is superimposed on the data Dl of the next word Wl to become the true value Pt, but this true value Pl also exceeds the full scale FS (in the negative direction) and overflows. Therefore, the output value Ql is clipped to the negative maximum value. This is how errors propagate. In addition, in FIG. 11, for convenience of illustration, the error-superimposed true value PO of the first word WO of the block is within several times the full scale, but in reality, the WO
The error fed back can be approximately 1000 times full scale (approximately 60 dB), and error propagation can persist for long periods of time.

On the other hand, if the amount of movement when the ranging position (requantization hit extraction position) moves toward the LSB side is limited to, for example, 1 bit as described above, the block The error fed back in the first word WO is kept small, and error propagation is eliminated in a short time. Therefore, the distortion of the output signal from the decoder is suppressed to a low level, and in combination with the temporal masking effect, signal transmission without any audible problem is possible.

It should be noted that the present invention is not limited to the above-mentioned embodiments, but is also applicable to the sampling frequency of the input digital signal, the number of hits in one word, the number of words in one block, the highest order N and the number of types of filters, Alternatively, it goes without saying that the number of requantization bits and the like can be set arbitrarily.

〔Effect of the invention〕

According to the signal transmission device of the present invention, when the requantization bit overflows, clipping is performed at the maximum positive or negative value, and the error at this time is fed back to perform noise shaping processing, thereby preventing error propagation. It is possible to effectively prevent the negative effects caused by the generation of abnormal noise.

In addition, prior to such ripping processing, the ranging position during requantization is set to the maximum absolute value within the block.
The above clipping process becomes more effective by performing it based on a value multiplied by the above coefficient γ or by limiting the amount of movement when the ranging position moves toward the LSB side.

That is, the peak value (maximum absolute value) in the block of output from the selected filter is multiplied by a coefficient γ (γ≧1), and the ranging position, that is, the requantization bit extraction position is determined based on this multiplied value. By determining , overflow is less likely to occur even if a feedback error due to noise shaping processing is superimposed.

In addition, when the input signal level suddenly decreases near the block boundary and the ranging position, that is, the requantization bit extraction position attempts to rapidly move toward the LSB side,
By limiting this amount of movement, errors caused by overflow of requantization bits can be suppressed.

Further, according to the signal transmission device of the embodiment of the present invention, a plurality of difference processing filters that output each take of high-order difference PCM, first-order difference PCM, and straight PCM are used, and these are adaptively switched and selected. Therefore, it is possible to efficiently reduce the bit rate, and it is possible to transmit signals at extremely low bit rates without deteriorating the signal quality. Furthermore, since the straight PcM data output mode can be switched and selected, it is possible to solve problems such as S/N deterioration when inputting a high-frequency signal and generation of excessive error power when an error occurs.

[Brief explanation of drawings]

FIG. 1 is a block circuit diagram showing a schematic configuration of the entire system to which the signal transmission device according to the present invention is applied, and FIGS. 2 and 3 show more specific configuration examples of the encoder and decoder shown in FIG. 4 is a graph showing the frequency characteristics of a plurality of differential processing filters, and FIG. 5 is a diagram showing an example of the transmission word structure within one block.
Fig. 6 is a graph showing the frequency characteristics of multiple differential processing filters used in other specific examples, Fig. 7 is a graph showing the spectral distribution of noise subjected to noise shaping processing, and Fig. 8 is a graph showing the spectral distribution of noise during requantization. Figure 9 is a block circuit diagram showing the main parts of the encoder, Figure 10 is a block circuit diagram showing the main parts of the decoder, and Figures 11 and 12 are diagrams for explaining the movement of the ranging position. FIG. 3 is a diagram for explaining error propagation due to overflow during conversion. 10・・・・・・・・・・・・・・・・・・・・・
・Encoders 12.12A to 12D, 20.34...
・・・Predictor 15.19.32・・・Shifter 16・・・・・・・・・・・・・・・・・・・・・・・・
・Quantizer 17・・・・・・・・・・・・・・・・・・
...Noise shaper 21...
・・・・・・・・・・・・Prediction/range adaptation circuit 26・・・・・・・・・・・・・・・・・・・・・
Clipping circuit thousand

Claims (1)

[Claims] A means for dividing an input digital signal into blocks of a certain number of words along the time axis, performing predictive processing on the signal for each block, requantizing the predictively processed signal, and feeding back the quantization error. means for performing noise shaping processing on the predicted signal; means for determining a requantization bit extraction position in the requantization based on the maximum absolute value within the block of the predicted signal; and the requantization data. and means for clipping to a maximum positive or negative value when an overflow occurs.