JPH0854899A - Encoding device - Google Patents

Abstract

PURPOSE:To improve the quantizing efficiency of data by optimizing the quantization of a scaling factor. CONSTITUTION:A scaling factor quantizing.encoding section 5 obtains a maximum quantization value FQ' which does not exceed a scaling factor F based on the factor F in the corresponding band range from a scaling factor extracting section 4 and a bit distribution number B from a bit distribution section 3. If the code of normalized values does not exceed the code assigned range determined by the number B during the normalization of the factor F by its quantized value FQ, FQ' is made to be the scaling factor quantized value FQ. Moreover, if FQ' exceeds the code assigned range, (FQ'+dFQ) (where dFQ is the quantization width of a scaling factor) is made to be the scaling factor quantized value FQ. Thus, a scaling factor is optimally quantized so that the input range of a quantizer in a quantizing-encoding section 7 is effectively used.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coding apparatus for normalizing digital data such as tone data or voice data by a scaling factor and coding the digital data.

[0002]

2. Description of the Related Art Musical sounds and voices have a wide dynamic range and a wide frequency range. Therefore, when digitized by PCM (pulse coding) with high sound quality, the bit length is 16 and is 44.1 kHz or 48 kHz. It is common to sample. The bit rate at this time is high, 70
5.6kbps (44.1kHz) or 768kbps (48kHz)
Is. Therefore, there is an encoding device that compresses the PCM digital data with high efficiency in order to reduce the bit rate to about 1/4 to 1/8.

In this encoding apparatus, in order to perform encoding without lowering the dynamic range of the input data and without degrading the precision in quantization at any amplitude level, the absolute value of the input data is calculated in frame units for a predetermined time. Find the maximum value,
The maximum value is quantized as a scaling factor.
Then, the input data in the frame is divided by the quantized scaling factor to normalize all the data to an absolute value of 1 (or a specific value) or less. In this way, by normalizing, the frequency component with a small level can be treated in the same manner as the frequency component with a large level, which is particularly effective for a band division type coding system in which the level difference between bands is large. The above normalization is performed for each frame and for each frequency band. A typical example of this band division type encoding system is MPEG1 / Audio (formally ISO / IEC11172-3) which is an international standard. The number of band divisions is constant.

As described above, in the above coding apparatus, since the input data is divided by the quantized scaling factor to normalize the data to a numerical value of "1" or less, the range of quantization of the scaling factor is It needs more than the dynamic range of the data. Therefore, in the quantization of the scaling factor, it is general to search for the smallest quantized value that exceeds the scaling factor. As a result, the normalized input data can always have a numerical value of "1" or less.

In general, quantization of this scaling factor is performed at equal intervals in decibel conversion. In MPEG1 / Audio, the quantization range of the scaling factor is 126 dB for each band, and 6 bits are assigned to the number of quantization bits.

[0006]

However, the conventional coding apparatus for quantizing the scaling factor has the following problems. That is, when the normalized input data is quantized by the number of bits allocated to the quantizer, the smaller the maximum absolute value of the normalized input data is, the smaller the value is (1) (that is, the scaling factor itself). (The larger the quantization error is), the quantization is not performed effectively, and the error becomes large. A specific example of this is shown in FIG.

FIG. 10 shows the quantization characteristics of a seven-stage (quantization bit number = 3) mid-tread type quantizer (which is called because there is a tread surface of a stairway at the origin). . The horizontal axis represents the amplitude of the input signal, and the vertical axis represents the amplitude after quantization. A code (3-bit binary number expression) is assigned to each tread. In the above-mentioned conventional coding apparatus, the smallest quantized value that exceeds the scaling factor is used as the scaling factor quantized value. Therefore, the quantized error of the scaling factor itself may be as large as -1.9 dB (0.804). Occurs. By the way, the quantization range when a signal such that the quantization error of the scaling factor itself is as large as -1.9 dB (0.804) is input as the input of the quantizer is as shown by the broken line in the figure. It becomes a rectangular area (a) surrounded by.

This quantization range means that the quantized value obtained by quantizing the input signal is within the normalization range, and
Amplitude of input signal (−7/6 to 7/6) and amplitude of quantized value (−1 to 1) that can assign codes to quantized values
It is a range within the range enclosed by and (hereinafter referred to as the code allocation range). Therefore, as mentioned above,
Quantization error of scaling factor itself is -1.9 dB
When a signal that becomes as large as (0.804) is input to a seven-stage (quantization bit number = 3) midtread quantizer, the maximum stage to which the code “011” is assigned and The smallest stage assigned the code "101" is not used, and this quantizer operates substantially as a 5-stage quantizer.

In this case, it cannot be said that the quantization is effectively performed, and coarse quantization will be performed. Therefore, the ratio of the quantization width to the amplitude of the input signal becomes relatively large. Since the maximum value of the data quantization error is 1/2 of the quantization width of the quantizer, the data quantization error also becomes large.

By the way, conventionally, the quantization of the scaling factor and the quantization of the data are performed independently. Therefore, when the data is quantized, the range of the input waveform becomes narrower than the code allocation range (that is, the original capability range of the quantizer) determined by the number of quantization bits, as shown by the area (a) in FIG. However, no measures are taken.

Therefore, an object of the present invention is to improve the sound quality of compressed acoustic data by optimizing the quantization of the scaling factor to improve the data quantization efficiency and suppress the quantization error. To provide a device.

[0012]

To achieve the above object, the invention according to claim 1 quantizes and encodes the scaling factor extracted by the scaling factor extraction unit by the scaling factor quantization / encoding unit, The normalization unit normalizes the input data using the obtained scaling factor quantized value, and quantizes the normalized value of this input data.
In an encoding device that quantizes and encodes in an encoding unit,
The scaling factor extraction unit extracts a maximum value of absolute values of a predetermined number of input data in one frame as a scaling factor, and the scaling factor quantization / encoding unit does not exceed the value of the scaling factor. Quantization / encoding means that obtains a quantized value and a minimum quantized value that exceeds the scaling factor, and that outputs a maximum quantized value that does not exceed the obtained scaling factor value as a scaling factor quantized value candidate, In a predetermined procedure, it is determined whether or not the scaling factor quantized value candidates exceed the range in which the code determined by the number of quantization bits in the quantization / encoding unit can be assigned. If it is determined that the value does not exceed While output as of value, if the candidate is determined to exceed the above range, the quantization and
A command signal for instructing the encoding means to calculate the minimum quantization value exceeding the scaling factor is output, and the minimum quantization value exceeding the scaling factor obtained by the quantization / encoding means is set to the scaling factor quantum. It is characterized in that it has a determination means for outputting the converted value.

According to a second aspect of the present invention, the scaling factor extracted by the scaling factor extraction unit is quantized and encoded by the scaling factor quantization / encoding unit, and the obtained scaling factor quantized value is used. In a coding device that normalizes input data in a normalization unit and quantizes and codes a normalized value of this input data in a quantization / coding unit, the scaling factor extraction unit is configured to input a predetermined number of inputs in one frame. The maximum value of the absolute value of the data is extracted as a scaling factor, and the threshold value in the range in which the code determined by the quantization bit number in the quantization / encoding unit can be assigned and the quantization bit number are associated with each other. The scaling factor quantization / encoding unit includes a table storage unit for storing the table The maximum quantized value that does not exceed the value of the ring factor and the minimum quantized value that exceeds the above scaling factor are obtained, and the maximum quantized value that does not exceed the value of the obtained scaling factor is output as a candidate for the quantized scaling factor value. With reference to the above table, it is determined whether or not the scaling factor quantized value candidates exceed the range in which the code determined by the quantizing / encoding means and the quantizing bit number in the quantizing / encoding unit can be assigned. If it is determined that the candidate does not exceed the above range, the candidate is output as a scaling factor quantized value, while if it is determined that the candidate exceeds the above range, the quantization or Outputs a command signal for instructing the encoding means to calculate the minimum quantized value exceeding the above scaling factor. It is characterized by having a determining means for outputting the minimum quantization value exceeding the scaling factor obtained by the quantization and encoding means as the scaling factor quantization value.

According to a third aspect of the present invention, the scaling factor extracted by the scaling factor extraction unit is quantized and encoded by the scaling factor quantization / encoding unit, and the obtained scaling factor quantized value is used. In the encoding device that normalizes the input data in the normalization unit and quantizes and encodes the normalized value of the input data in the quantization / encoding unit, the scaling factor extraction unit includes a predetermined number of frames in one frame. It is a threshold value in a range in which the maximum absolute value of input data is extracted as a scaling factor and a code determined by the number of quantization bits in the quantization / encoding unit can be assigned to the quantized value of the scaling factor. Make sure that the value of the scaling factor and the second threshold value obtained by adding the first threshold value do not exceed the second threshold value. A table storing unit for storing the minimum quantized value and the code corresponding to each other is provided for each of the quantized bit numbers, and the scaling factor quantization / encoding unit refers to the table. To retrieve the minimum second threshold value exceeding the scaling factor, and to retrieve the second retrieved threshold value.
It is characterized by having a quantized value search means for outputting a quantized value associated with a threshold as a scaling factor quantized value.

[0015]

According to the first aspect of the present invention, the scaling factor extraction unit extracts the maximum absolute value of a predetermined number of input data in one frame as the scaling factor. Then, the quantizing / encoding means of the scaling factor quantizing / encoding unit obtains the maximum quantized value that does not exceed the above scaling factor, and outputs it as a candidate for the quantized scaling factor. Then, the determination means in the scaling factor quantization / encoding unit assigns a code in which the candidates for the scaling factor quantization value from the quantization / encoding unit are determined by the number of quantization bits in the quantization / encoding unit. Is determined by a predetermined procedure. When it is determined that the candidate does not exceed the range, the candidate is output as the scaling factor quantized value.

On the other hand, when it is determined that the candidate exceeds the range, a command signal for instructing the quantization / encoding means to calculate the minimum quantized value exceeding the scaling factor is output. Then, the minimum quantized value exceeding the scaling factor obtained by the quantizing / encoding means is output as the scaling factor quantized value.

Thereafter, the normalization unit normalizes the input data using the scaling factor quantized value output from the determining means, and the normalization value of the input data is quantized by the quantization / encoding unit. And encoded. At this time, the quantized value obtained by quantizing the input data by the quantizing / encoding unit is within the widest quantized range within which the code can be assigned in the quantizing / encoding unit. Fits in.

According to the second aspect of the invention, the scaling factor extraction unit extracts the maximum absolute value of a predetermined number of input data in one frame as the scaling factor. Then, the quantizing / encoding means of the scaling factor quantizing / encoding unit obtains the maximum quantized value that does not exceed the above scaling factor, and outputs it as a candidate for the quantized scaling factor. Then, the scaling factor quantization
Whether the table stored in the table storage unit is referred to by the determination unit in the encoding unit and whether the above candidate exceeds a range in which the code determined by the number of quantization bits in the quantization / encoding unit can be assigned. Is determined. When it is determined that the candidate does not exceed the range, the candidate is output as the scaling factor quantized value.

On the other hand, when it is determined that the candidate exceeds the range, a command signal for instructing the quantization / encoding means to calculate the minimum quantized value exceeding the scaling factor is output. Then, the minimum quantized value exceeding the scaling factor obtained by the quantizing / encoding means is output as the scaling factor quantized value.

According to the third aspect of the invention, the scaling factor extraction unit extracts the maximum absolute value of a predetermined number of input data in one frame as the scaling factor. Then, the quantization value search means in the scaling factor quantization / encoding unit refers to the table stored in the table storage unit to search for the minimum second threshold value exceeding the scaling factor, and the second threshold value. The quantized value associated with the threshold value is obtained. Then, the obtained quantized value is output as the scaling factor quantized value.

[0021]

The present invention will be described in detail below with reference to the embodiments shown in the drawings. In the present embodiment, a band division type coding device will be described as an example, but the present invention can also be applied to a coding device using another scaling factor.

<First Embodiment> FIG. 1 is a block diagram showing an embodiment of an encoding apparatus according to the present embodiment. Digital tone data and voice data (hereinafter simply referred to as input data)
Is input to the input terminal 1 and becomes data of N sets of bands by the band division unit 2 including N band pass filters.
Each band is represented by a numerical value of 1 ≦ n ≦ N. The number of bits of the input data is 16 bits. However, in FIG. 1, the band “1” and the band “n” are shown for simplification.
It is represented by three bands of "," and "N".

Of the N sets of band data thus obtained, the band n data are grouped in frame units (units of I samples in terms of the number of data), and the scaling factor extraction unit 4 _n outputs the data D _n ( i) The maximum value f _n of absolute values of (0 ≦ i ≦ I−1) is extracted, and this maximum value f _n is used as a scaling factor. The decibel value of the scaling factor f _n is represented by F _n (= 20log ₁₀ f _n ). The bit allocation unit 3 determines the allocation of the number of bits when quantizing data for each band as follows. That is, FFT
(Fast Fourier transform) and the power of each band are used to analyze the power distribution of the input data in the frequency domain, apply human auditory characteristics (especially masking characteristics), and quantize error for each band. Compute the number of bits needed to be masked. The bit allocation number B _n of the band n obtained by the bit allocation unit 3 in this way is sent to the scaling factor quantization / encoding unit 5 _{n of} the band n, the quantization / encoding unit 7 _{n of the} band n, and the multiplexer 8. It Although the input data is directly input to the bit allocation unit 3 in FIG. 1, the output data for each band in the band division unit 2 may be input.

[0024] The scaling factor quantization and coding unit 5 _n is quantized scaling factors F _n on the basis of the bit allocation number B _n from the scaling factor F _n and the bit allocation portion 3 from the scaling factor extracting unit 4 _n The obtained scaling factor quantized value FQ _n and its code FC _n are sent to the multiplexer 8. Also, the scaling factor quantized value FQ _n is sent to the normalization unit 6 _n . The operation of the scaling factor quantizer / encoder 5 _n , which is a feature of this embodiment, will be described in detail later.

In the normalizing unit 6 _n , the data D _n (i) of the band n obtained by the band dividing unit 2 is divided by the true value fQ _n (= 10 ^{FQn / 20} ) of the scaling factor quantized value FQ _n. Then “±
Normalize to a number N _n (i) within the range of 1 ", and
This normalized data N _n (i) is quantized and encoded by the quantization / encoding unit 7 _n . At this time, the quantizing / encoding unit 7 _n has a midtread type quantizer similar to that shown in FIG. 10, and this quantizer responds to the bit allocation number B _n given by the bit allocation unit 3. The number of stages is “2 ^Bn −1” and the quantization width is dQ _n
Is a ^{_{dQ n = 1 / {2 (}} Bn-1) -1} ... (1).

The quantization / encoding unit 7 _n quantizes and encodes the normalized data N _n (i) with the quantization width dQ _n ,
The code C _n is obtained. In _{C n (i) = round (} N n (i) / dQ n) ... (2) where, "round english (us) (x)" is a function that integer to x by rounding the first decimal place of x Is. still,
The code C _n (i) is expressed by the bit allocation number B _n and is sent to the multiplexer 8.

The multiplexer 8 collects the code C _n (i) of the data D _n (i) of each band n, the bit allocation number B _n and the code FC _{n of the} scaling factor in frame units according to a certain format. Into a bitstream. This bit stream is transmitted from the output terminal 9 to the outside.

Next, the scaling factor quantization
Quantization and coding operations by the coding unit 5 _n will be described in detail. Scaling factor quantization / encoding unit 5
_In general, _n has only quantization / coding means, and as described above, the scaling factor F _n is quantized by searching for the minimum scaling factor quantized value FQ _n that exceeds the scaling factor F _n . Turn into. That is, F _n ≦ FQ _n = Fl + FC _n × dFQ (0 ≦ FC _n ≦ 2 ^SB −1) (3) Here, Fl: lower limit value of scaling factor quantized value dFQ: quantization width of scaling factor F _n SB: is to explore the minimum FQ _n and FC _n satisfying the number of quantization bits of the scaling factor F _n.
The expression (3) is a decibel dimension.

However, in this embodiment, the scaling factor F _n is quantized as follows. FIG. 2 is a detailed block diagram of the scaling factor quantizer / encoder 5 _n in this embodiment. Quantization
The encoding means 41 _n searches for maximum FQ ′ _n and FC ′ _n that satisfy F _n ≧ FQ ′ _n = Fl + FC ′ _n × dFQ (0 ≦ FC ′ _n ≦ 2 ^SB −1) (4). Alternatively, determine the scaling factor quantization values FQ _n and its sign FC _n based on similarly to the conventional equation _{(3), FQ 'n =} FQ n -dFQ FC' n = FQ by FC _n -1 _'n and FC 'You may ask for _n . The quantized value FQ ′ _n and its code FC ′ _n thus obtained are the scaling factor quantized value FQ _n and its code FC _n.
Is sent to the determination means 42 _n as a candidate for

The determining means 42 _n normalizes the true value f _n of the scaling factor F _n with the true value fQ ' _n (= 10 ^{FQ'n / 20} ) of the candidate scaling factor quantized value FQ' _n . The divided value (which is also the quantization error of the scaling factor F _n ) is encoded according to the equation (2) (however, the normalized data N _n (i) in the equation (2) is f _n / fQ. 'replace with _n ). Code thus obtained is greater than or equal to a maximum value 2 ^(Bn-1) of the assignable codes determined by the bit allocation number B _n. As a result, it does not exceed (i.e., the code assignment does not exceed the range) be determined that, to determine the quantized value FQ _'n and a scaling factor quantization value FQ _n. Also, it is to determine the FC _'n the code FC _n scaling factor F _n. Then, the determined scaling factor quantized value FQ _n and the code FC _n of the scaling factor F _n are output.

However, if it is determined that the code obtained by encoding the normalized value of the scaling factor F _n according to the equation (2) exceeds the maximum value 2 ^(Bn-1) , the quantization / encoding is performed. A command signal for commanding calculation of the minimum quantized value FQ _n and its code FC _n that satisfy the equation (3) is output to the means 41 _n . Then, the minimum quantized value (that is, (FQ ' _n + dFQ)) that exceeds the scaling factor F _n sent from the quantization / encoding means 41 _n and its code (that is, (FC' _n +1)) Is determined as the scaling factor quantized value FQ _n and its code FC _n . And
The determined scaling factor quantized value FQ _n and the code FC _n of the scaling factor F _n are output.

That is, the scaling factor quantization / encoding unit 5 _n in this embodiment has a scaling factor smaller than the quantization value obtained by the normal quantization of the scaling factor by the quantization width dFQ of the data D _n (i). When the quantization value is selected by the quantization / encoding means 41 _n , it is determined whether the scaling factor quantization value exceeds the code allocation range defined by the bit allocation number B _n.
The scaling factor quantized value FQ _n is optimally selected by examining _n .

The above will be described in detail with reference to FIG. Now, the quantization width dF of the scaling factor F _n
When Q is set to “2 dB”, the quantization error in the conventional quantization of the scaling factor according to the equation (3) is −1.9 dB.
(0.804). First, as mentioned above,
By (4), a quantized value smaller than the conventional scaling factor quantized value by the quantization width dFQ is selected as the scaling factor quantized value candidate FQ ′ _n . Then, the quantization error of the scaling factor F _n is _{equal to} the conventional quantization error (-1.9 dB (0.804)) and the quantization width dF.
The value obtained by adding Q is −1.9 dB (0.804) +2 dB = 0.1 dB (1.012). Therefore, the absolute value of the normalized value of the scaling factor F _n will exceed "1".

However, this scaling factor F _n
The absolute value "1.012" of the normalized value of is smaller than the boundary "7/6" of the input value where the quantized output value exceeds "1" (the stage to which the maximum code is assigned) (in other words, the above code since over not) the allocation range, the candidate FQ _'n quantized value of the scaling factor F _n is determined a scaling factor quantization value FQ _n. The quantization range at that time is the rectangular area surrounded by the alternate long and short dash line in FIG.
It is (b). In this case, the conventional quantization range (rectangular area
Since it is wider than (a)) by the quantization width dFQ, the maximum stage to which the code “011” is allocated and the minimum stage to which the code “101” is allocated are also used. The quantization error is smaller than that in the past.

The absolute value of the normalized value of the scaling factor F _n is larger than the boundary "7/6" of the input values whose quantized output value exceeds "1" (in other words, it exceeds the code allocation range). ) in this case, one step from the candidate FQ _'n quantized value of the scaling factor F _n greater (i.e., plus quantization width dFQ scaling factors F _n) value determined a scaling factor quantization value FQ _n the Of. The quantization range at that time becomes the conventional quantization range.

Data of other bands output from the band division unit 2 are processed in exactly the same manner as the case of the data of the band n.

Next, the decoding will be described. FIG.
It is a block diagram which shows one Example of the decoding apparatus which decodes the code | cord obtained by the encoding apparatus of a present Example. FIG.
The bit stream output from the output terminal 9 of the encoding device shown in FIG.
(There is a semiconductor memory or the like) through the input terminal 11 of the decoding device.

From the bit stream input to the input terminal 11, the demultiplexer 12 outputs each band n.
Separately, the code C _n (i), the bit allocation number B _n, and the scaling factor code FC _n are extracted in frame units. The scaling factor code FC _n is input to the scaling factor decoding / inverse quantization unit 13 _n . Then, the scaling factor decoding / inverse quantization unit 13 _n
The scaling factor quantized value FQ _n is obtained by the equation (3) and sent to the denormalization unit 15 _n . The decoding / dequantization unit 14 _n performs an operation reverse to that of the quantization / encoding unit 7 _n in the above-described encoding device, based on the input code C _n (i) and the bit allocation number B _n . Dequantized data NQ _n
Find (i). Specifically, using the bit allocation number B _n ,
The quantization width dQ _n is obtained by the equation (1), and the obtained quantization width dQ _n is dequantized by the equation (5) to calculate the data NQ _n (i). NQ _n (i) = C _n (i) × dQ _n (5)

The data NQ _n (i) thus obtained remains in the normalized state. Therefore, the data is sent to the denormalization unit 15 _n , and the denormalization unit 15 _n multiplies the scaling factor quantized value FQ _{n by} the true value fQ _n to obtain the data DQ _n (i) in the band n. The data DQ ₁ (i) to DQ _N (i) (0 ≦ i ≦ I-1) of the entire band thus obtained are combined by the band synthesizing unit 16 and restored to digital musical sound data or voice data and output. It is output from the terminal 17.

As described above, in the present embodiment, the scaling factor extraction unit 4 _n in the above encoding device
The maximum absolute value of the data D _n (i) is extracted as the scaling factor F _n for each band. The quantization / encoding means 4 of the scaling factor quantization / encoding unit 5 _n is also used.
1 _n by searching the maximum scaling factor quantization value FQ _'n and the code FC' _n satisfying the equation (4). Then, the determination unit 42 _n, exact numeric f _n of the scaling factor F _n from scaling factor extracting unit 4 _n
Is normalized by the exact value fQ ' _n of the quantized value FQ' _n , and is encoded according to the equation (2). Then, if it is determined that the obtained code does not exceed the code assignment range determined by the bit allocation number B _n, the quantization value FQ 'a _n and a scaling factor quantization value FQ _n, the FC' scaling _n the code FC _n factors F _n.

On the other hand, when it is determined that the code when the normalized value of the exact value f _n of the scaling factor F _n is encoded according to the equation (2) exceeds the code allocation range, quantization / encoding is performed. minimum quantization value satisfying the equation (3) by means _{41 n (FQ 'n + dFQ} ) one of a scaling factor quantization value FQ _n seek, the code FC _n scaling factor F _n (FC' _n +1) And

That is, in the present embodiment, when the largest quantized value that does not exceed the scaling factor F _n is selected as the scaling factor quantized value candidate, that candidate is quantized / coded by the bit allocation number B _n . Conversion part 7 _n
If it does not exceed the code allocation range in, the candidate is set as the scaling factor quantized value FQ _n . On the other hand, when it exceeds, the smallest quantized value exceeding the scaling factor F _n is set as the scaling factor quantized value FQ _n . By doing so, when quantizing the normalized data N _n (i) that is normalized with the scaling factor quantized value FQ _n , the obtained quantized value is the widest quantized range that does not exceed the code allocation range. The scaling factor F _n can be quantized to fit within. Therefore, the quantization of the data D _n (i) is effectively performed, the quantization error is reduced, and the encoded and decoded data with higher sound quality can be obtained.

The quantizer incorporated in the quantizer / encoder 7 _n in the above embodiment is a mid-treader quantizer as shown in FIG. The midtread type quantizer has a uniform quantization width and is linear (uniform).
However, the quantizer in this invention is non-midtread, non-linear, and non-uniform,
The maximum quantized value FQ ′ _n that satisfies the equation (4) is the bit allocation number B
If it is possible to determine whether or not the code allocation range defined by _n is exceeded, the same as described above can be performed.

<Second Embodiment> This embodiment is an embodiment in which the same effect as that of the first embodiment is obtained by using a table. That is, the threshold value TQ of the code allocation range determined by the bit allocation number B _n is obtained in advance and tabulated. And
By referring to the table, it is determined whether or not the quantized value exceeds the code allocation range when the scaling factor F _n is quantized.

Similar to the first embodiment, the midtread quantizer will be specifically described below as an example. FIG.
FIG. 3 is a block diagram of an encoding device according to the present embodiment.
The input terminal 21, the band division unit 22, the bit allocation unit 23, the scaling factor extraction unit 24 _n , the normalization unit 26 _n , the quantization / encoding unit 27 _n , the multiplexer 28 and the output terminal 29 in this embodiment are the first Input terminal 1 in the embodiment,
The band dividing unit 2, the bit allocating unit 3, the scaling factor extracting unit 4 _n , the normalizing unit 6 _n , the quantizing / encoding unit 7 _n , the multiplexer 8 and the output terminal 9 are exactly the same, and the description thereof will be omitted.

The above-mentioned encoding device has a table storage unit 30 for storing the above-mentioned table, and the scaling factor quantization / encoding unit 25 _n causes the scaling factor F _n of the scaling factor F _n to be described in detail below. It quantizes and encodes.

In the midtread quantizer, the threshold value tq of the code allocation range determined by the bit allocation number B _n is tq = (2 ^B -1) /, using the quantization bit number B of data. It can be expressed as (2 ^B -2) ... (6). For example, as shown in FIG. 10, when the quantization bit number B is “3”, tq = 7/6. Therefore, the quantization bit number B and the threshold value tq calculated based on the equation (6)
Make a table of the relationship with. At that time, the threshold value tq is set to a decibel value TQ. Then, similarly to the first embodiment, the scaling factor quantization value true numerical f _n of the scaling factor F _n candidate FQ _n in the normalized values f _n / fQ _'n' exact numeric fQ of _n ' decibel expression (F _n -FQ _'n)
The scaling factor quantized value is determined by determining whether (which is also the quantization error of the scaling factor F _n ) exceeds the threshold TQ.

Here, in the present embodiment, f _n / fQ ' _n
In order to simplify the configuration of the scaling factor quantization / encoding unit 25 _n by eliminating the calculation processing of F _n / fQ ′ _n by an exact number expression, the quantization error (F _n −F
Instead of Q ′ _n ), the scaling factor quantization / encoding unit 25 _n calculates a quantization error (F _n −FQ ′ _n ). In response to this, the table is stored in the table storage unit 30 in decibel representation as shown in FIG. In FIG. 6, the range of the quantization bit number B is 2 ≦ B ≦ 16.

The encoding device having the above configuration operates as follows. FIG. 5 is a detailed block diagram of the scaling factor quantizer / encoder 25 _n in this embodiment. The quantizing / encoding means 51 _n has the maximum quantized value FQ ′ satisfying the expression (4) based on the scaling factor F _n extracted by the scaling factor extracting unit 24 _n , as in the first embodiment. Search for _n and its code FC ' _n . Then, it sends a 'the code FC and _{n' n} candidates FQ of the scaling factor quantization values FQ _n to the determining means 52 _n.

[0050] The determination means 52 _n, based on the scaling factor F _n from candidates FQ _'n and a scaling factor extractor 24 _n from the quantization and coding unit 51 _n, the quantization of the scaling factor F _n error (F _n -FQ '
_n ) is calculated. Then, the table (see FIG. 6) stored in the table storage unit 30 is searched based on the bit allocation number B _n distributed from the bit allocation unit 23 to obtain the threshold TQ _n of the code allocation range in the band n. . Then, 'examined the magnitude relation _(n and threshold TQ _n, F _n -fq quantization error scaling factor F _n, which is the calculated F _n -fq)' if _n <TQ _n, data D _n ( since the quantized values also quantize the maximum value of the absolute values of i) it does not exceed the code assignment range, to accept the candidate FQ _'n scaling factors quantized values and scaling factor quantization value FQ _n output To do. On the other hand, if F _n −FQ ′ _n ≧ TQ _n , the quantized value exceeds the code allocation range even if the maximum absolute value of the data D _n (i) is quantized.
A command signal for instructing the calculation of the minimum quantized value FQ _n and its code FC _n that satisfy the equation (3) is output to the quantization / encoding means 51 _n . Then, the scaling factor quantized value FQ _n is output as (FQ ′ _n + dFQ) based on the calculated value sent from the quantization / encoding means 51 _n .

As described above, in the above embodiment, the determining means 52 of the scale factor quantizing / encoding unit 25 _n.
n Since candidate FQ _'n of skating phosphide factor quantization value performed by referring to the table of the determination process whether exceeds the above code assignment range, it is possible to easily perform the determination process. Therefore, the configuration of the determination means 52 _n can be simplified.

<Third Embodiment> This embodiment is another embodiment in which the same effect as that of the first embodiment is obtained by using a table.
FIG. 7 is a block diagram of an encoding unit in the encoding device according to the present embodiment. The input terminal 31, in the present embodiment,
Band division unit 32, bit allocation unit 33, scaling factor extraction unit 34 _n , normalization unit 36 _n , quantization / encoding unit 37 _n ,
The multiplexer 38 and the output terminal 39 are the input terminal 1, the band division unit 2, the bit allocation unit 3, the scaling factor extraction unit 4 _n , the normalization unit 6 _n , the quantization / encoding unit 7 _n , and the multiplexer in the first embodiment. 8 and the output terminal 9 are exactly the same. Further, the table storage unit 40 in the present embodiment.
The table stored in is the following table.

The equation (7) is obtained by adding the threshold TQ (hereinafter referred to as the first threshold) of the code allocation range represented by the equation (6) to the equation (3). _{F n <FQ n + TQ n} = Fl + FC n × dFQ + TQ n = TF n ... (7) Then, the minimum scaling factor quantization value does not exceed the second threshold value TF _n in the scaling factor F _n is Equation (7) FQ _n If the code FC _n is searched, the data D
the quantized value when the maximum value of the absolute value and quantization of _n (i) is able quantization scaling factor F _n so as not to exceed the code allocation range.

Therefore, in this embodiment, the minimum scaling factor quantized value FQ _n that satisfies the equation (7) and its code FC _n and the second threshold value TF _{n at} that time are associated with each other and the bit allocation number B _n is associated. Each table is stored in the table storage unit 40. FIG. 9 shows an example of the above table, where Fl = −30 dB, dFQ = 2 dB, SB = 6 bits, and the scaling factor quantized value FQ and its code F when the data bit allocation number B = 3.
The relationship between C and the second threshold value TF is shown. However, the subscript "n" indicating the band is omitted.

The scaling factor quantization / encoding 35 _n in the encoding device having the above-mentioned configuration operates as follows. FIG. 8 is a detailed block diagram of the scaling factor quantizer / encoder 35 _n . Here, it is assumed that the bit allocation number B _n of the band n from the bit allocation unit 33 is “3”. Then, the quantized value search means 61 _n has a table corresponding to the scaling factor F _n calculated by the scaling factor extraction unit 34 _n and the bit allocation number B _n = 3 stored in the table storage unit 40 (see FIG. 9). Second
Comparing the threshold value TF _n, greater than the scaling factor F _n, and searches the nearest second threshold TF _n.

For example, _assuming that the scaling factor F _n = 2.1 dB obtained above, the second threshold value TF _n = 3.
339 dB is obtained. At that time, FQ _n = in the same row as the second threshold value TF _n = 3.339 dB in the table.
2 dB and FC _n = 16 are the target scaling factor quantized value FQ _n and its code FC _n . Therefore, the quantized value retrieving means 61 _n outputs the scaling factor quantized value FQ _n thus obtained and its code FC _n .

In the present embodiment, the scaling factor quantizing / encoding unit 35 _n uses the quantized value search means 61.
_The scaling factor quantized value FQ _n that does not exceed the code allocation range when the maximum absolute value of the data D _n (i) is quantized by simply drawing the table stored in the table storage unit 40 by _n. And its code FC _n can be obtained. Therefore, the scaling factor quantization / encoding unit 35 _n does not require arithmetic processing, and the processing of the scaling factor quantization / encoding unit 5 _n in the first embodiment and the scaling factor quantization / encoding unit in the second embodiment.
Compared with the processing of the encoding unit 25 _n, the objective can be achieved with less processing. Therefore, it is possible to simplify the configuration of the scale factor quantization / encoding unit 35 _n .

However, since it is necessary to create a table as shown in FIG. 9 for each bit allocation number B _n and store it in the table storage unit 40, the table storage unit 40 has a capacity for storing a large table. is necessary. For example, the range of the bit allocation number B _n when quantizing the data D _n (i) is set to 2 ≦ B _n ≦ 16, and the quantization bit number SB when quantizing the scaling factor F _n is “ 6 ″, the second threshold value TF _n in the table storage unit 40 is 15 × 6.
It is necessary to store 4 = 960 pieces.

In the third embodiment, the scaling factor quantizing / encoding unit 25 _n includes the table storage unit 40.
The minimum scaling factor quantized value FQ _n and its code FC _n satisfying the expression (7) are obtained by drawing the table stored in However, the present invention is not limited to this, and the scaling factor quantizing / encoding unit 35 _n is configured as shown in FIG.
By directly calculating the equation (7), the minimum scaling factor quantized value FQ _n and its sign F that satisfy the equation (7) are obtained.
There is no problem in obtaining C _n .

[0060]

As is clear from the above, in the encoding device according to the first aspect of the invention, the scaling factor extraction unit extracts the maximum absolute value of the input data as a scaling factor, and the quantization / encoding means. If the maximum quantized value that does not exceed the value of the scaling factor is determined as a candidate by the determining means, and the determination means determines that the candidate does not exceed the range to which the code in the quantization / encoding unit can be assigned, While outputting the candidate as a scaling factor quantized value, if it is determined to exceed, output the minimum quantized value that exceeds the value of the scaling factor obtained by the quantizing / encoding means as a scaling factor quantized value. Therefore, when quantizing the input data normalized by the obtained scaling factor quantization value, Quantization values to fit within the broadest quantization range not exceeding the above range can be optimally quantize the scaling factor to be.

Therefore, according to the present invention, the quantization
It is possible to improve the quantization efficiency by reducing the relative ratio of the quantization width to the amplitude of the input data when the input unit quantizes the input data. As a result, the quantization error of the input data is suppressed and the sound quality when the acoustic data is compressed is improved.

Further, in the encoding device of the invention according to claim 2, the determining means refers to the table stored in the table storage section to determine the scaling factor quantized value obtained by the quantizing / encoding means. It is determined whether or not the candidate exceeds the range in which the code in the quantization / encoding unit can be assigned, and if it is determined that the code does not exceed, the candidate is output as a scaling factor quantized value, while it is determined to exceed it. In this case, the minimum quantized value that exceeds the value of the scaling factor obtained by the quantization / encoding means is output as the scaling factor quantized value, so that the optimum quantization of the above scaling factor is efficiently used by using a table. I can do it.

In the encoding device according to the third aspect of the present invention, the scaling factor extraction unit extracts the maximum absolute value of the input data as a scaling factor,
A first threshold value that is a threshold value in a range in which the quantization value search unit refers to the table stored in the table storage unit and can assign the code in the quantization / encoding unit to the quantization value of the scaling factor. Second value
When the threshold value is set, the minimum second threshold value that exceeds the scaling factor is searched, and the quantized value associated with the searched second threshold value is output as the scaling factor quantized value. The optimum quantization of factors can be performed simply by looking up the above table, and can be executed more efficiently.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an encoding device of the present invention.

FIG. 2 is a detailed block diagram of a scaling factor quantization / encoding unit in FIG.

3 is a block diagram showing an embodiment of a decoding device that decodes the code obtained by the coding device shown in FIG. 1. FIG.

FIG. 4 is a block diagram of an encoding device different from that in FIG.

5 is a detailed block diagram of a scaling factor quantizer / encoder in FIG. 4. FIG.

FIG. 6 is a diagram showing an example of a table stored in a table storage section in FIG.

FIG. 7 is a block diagram of an encoding device different from FIGS. 1 and 4.

8 is a detailed block diagram of a scaling factor quantizer / encoder in FIG. 7. FIG.

9 is a diagram showing an example of a table stored in a table storage section in FIG.

FIG. 10 is an explanatory diagram of a quantization range in a quantizer.

[Explanation of symbols]

2, 22, 32 ... Band division section, 3, 23, 33 ... Bit allocation section, 4, 24, 34 ... Scaling factor extraction section, 5, 25, 35 ... Scaling factor quantization / encoding section, 6, 26, 36 ... Normalization unit, 7, 27, 3
Reference numeral 7 ... Quantization / encoding section, 8, 28, 38 ... Multiplexer, 30, 40 ... Table storage section, 41, 51 ... Quantization / encoding means, 42, 52 ... Judging means, 61 ... Quantization value searching means .

Claims (3)

[Claims] 1. A scaling factor extracted by a scaling factor extraction unit is quantized and encoded by a scaling factor quantization / encoding unit, and an input data is normalized by a normalization unit using the obtained scaling factor quantized value. In the encoding device that converts the normalized value of the input data into a quantization / encoding unit and encodes the normalized value, the scaling factor extraction unit includes a maximum absolute value of a predetermined number of input data in one frame. And the scaling factor quantizing / encoding unit obtains a maximum quantized value that does not exceed the value of the scaling factor and a minimum quantized value that exceeds the scaling factor, and the obtained scaling factor. The maximum quantized value that does not exceed the value of Whether or not the scaling factor quantized value candidate exceeds the range in which the code determined by the quantization / encoding means which outputs as the candidate of the quantized value and the quantization bit number in the quantization / encoding unit can be assigned. If it is determined that the above candidate does not exceed the above range, the above candidate is output as a scaling factor quantized value, while if it is determined that the above candidate exceeds the above range, Outputting a command signal for instructing the quantization / encoding means to calculate a minimum quantization value that exceeds the scaling factor, and the minimum quantization exceeding the scaling factor obtained by the quantization / encoding means. An encoding device comprising: a determining unit that outputs a value as a scaling factor quantized value. 2. The scaling factor extracted by the scaling factor extraction unit is quantized and encoded by the scaling factor quantization / encoding unit, and the obtained scaling factor quantized value is used to normalize the input data by the normalization unit. In the encoding device that converts the normalized value of the input data into a quantization / encoding unit and encodes the normalized value, the scaling factor extraction unit includes a maximum absolute value of a predetermined number of input data in one frame. Is stored as a scaling factor, and a table storage that stores a table in which a threshold value in a range in which a code determined by the number of quantization bits in the quantization / encoding unit can be assigned and the number of quantization bits are associated with each other The scaling factor quantization / encoding unit is provided with a value of the scaling factor. Quantization / encoding that finds a maximum quantized value that does not exceed and a minimum quantized value that exceeds the scaling factor, and that outputs a maximum quantized value that does not exceed the value of the obtained scaling factor as a scaling factor quantized value candidate. Means, while determining with reference to the table whether the candidate of the scaling factor quantized value exceeds the range in which the code determined by the number of quantization bits in the quantization / encoding unit can be assigned, If it is determined that the candidate does not exceed the range, the candidate is output as a scaling factor quantized value, while if it is determined that the candidate exceeds the range, the quantization / encoding means Output a command signal for instructing the calculation of the minimum quantized value that exceeds the above scaling factor, Encoding apparatus characterized by comprising determination means for outputting the minimum quantization value exceeding the scaling factor obtained by the encoding means as a scaling factor quantization value. 3. The scaling factor extracted by the scaling factor extraction unit is quantized and encoded by the scaling factor quantization / encoding unit, and the normalization unit normalizes the input data using the obtained scaling factor quantized value. In the encoding device that converts the normalized value of the input data into a quantization / encoding unit and encodes the normalized value, the scaling factor extraction unit includes a maximum absolute value of a predetermined number of input data in one frame. Is extracted as a scaling factor, and a first threshold value that is a threshold value in a range in which a code determined by the number of quantization bits in the quantization / encoding unit can be assigned to the quantized value of the scaling factor is added. A second threshold and a minimum of the quantization that prevents the value of the scaling factor from exceeding the second threshold. And a table in which the codes are associated with each other are provided for each quantization bit number, and the scaling factor quantization / encoding unit refers to the table and stores the scaling factor. A code having a quantized value search means for searching the minimum second threshold value that is exceeded and outputting the quantized value associated with the searched second threshold value as a scaling factor quantized value. Device.