JP3383202B2 - Digital data decoding method and decoding device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital data decoding method and a decoding device for decoding encoded digital data by performing block floating processing.

[0002]

2. Description of the Related Art Conventionally, fixed-point DSP (digita)
When a digital input signal is encoded by using a digital signal processing device, a block floating technique is known in which the digital input signal is divided into blocks and then a floating process is performed on a block-by-block basis.

As this block floating technology,
When transforming spectrum data on the frequency axis using orthogonal transformation and encoding, by determining the block length and block floating coefficient using the same index such as the maximum absolute value of data before orthogonal transformation. , A proposal has been made to reduce the processing amount for determining the block length (Japanese Patent Laid-Open No. Hei 4-
No. 302540).

Further, at the time of the orthogonal transform and the inverse orthogonal transform calculation, the presence or absence of scale down is judged in each calculation process by comparing the input data value of each calculation formula with a preset scale down judgment reference value. A proposal has been made to reduce the calculation error due to the decrease in the number (Japanese Patent Laid-Open No. 6-164414).

According to this method, with respect to each calculation process of the inverse orthogonal transform, it is determined whether or not to scale down by looking at the maximum value every time, so that a small signal with a small spectrum is greatly scaled up and calculation accuracy is increased. Is improved.

In addition, the applicant of the present application is
In order to prevent overflow in the calculation process, we propose to further reduce the calculation error by performing fixed scale-down before each calculation process according to the maximum value of the input data before the inverse orthogonal transform calculation. (JP-A-7-366)
No. 66).

According to this method, before the inverse orthogonal transform calculation, the maximum margin bit number is determined in advance so as not to overflow, and the scale is fixedly scaled accordingly, so that the calculation accuracy of a large signal with a large spectrum is improved. It

[0008]

However, the digital signal decoding device disclosed in the above-mentioned Japanese Patent Laid-Open No. 6-164414 carries out scale-up or scale-down by observing an input data string in each calculation process. Therefore, it consumes a large amount of current and is not suitable for portable small devices.

Further, the digital signal processing device disclosed in the above-mentioned Japanese Patent Laid-Open No. 7-36666 was wasted in the initial shift amount before the inverse orthogonal transform calculation. Further, since it is necessary to obtain the maximum absolute value of the input data string before the inverse orthogonal transform calculation, the current is wasted.

The present invention has been made to solve the above problems, and its purpose is to determine the scale amount of block floating with a small amount of processing, and to perform scale-up and scale-down in more detail than 1 bit. It is an object of the present invention to provide a digital data decoding method and a decoding device which can be performed.

[0011]

In order to solve the above problems, a digital input signal is orthogonally transformed into spectrum data in the frequency domain for each unit time in order to solve the above problems. The data is divided into units in the frequency domain, and the spectrum data of each unit is quantized and encoded based on the representative value of the spectrum data of the unit to which it belongs. In a digital data decoding method for decoding coded data recorded in a format having at least a representative value of spectrum data, the maximum value of the representative value of the spectrum data of each unit is searched, and the maximum value is searched. On the basis of,
It is characterized in that the scale amount is optimally determined so that overflow does not occur in the process of inverse quantization and inverse orthogonal transform, and scale up and scale down are performed.

With the above arrangement, it is possible to eliminate the calculation error caused by the overflow. In addition, in order to obtain the block floating scale amount, it is sufficient to search for the representative value of the spectrum data of each unit, which has a smaller number than the spectrum data. Since this representative value is an absolute value, the block floating scale amount can be calculated. The required processing steps can be shortened.

Therefore, since the number of processing steps can be reduced as compared with the conventional case, it is possible to reduce the current consumption while eliminating the calculation error caused by the overflow.

In order to solve the above-mentioned problems, the method for decoding digital data according to a second aspect of the present invention, in addition to the configuration of the first aspect, performs the above scale-up and scale-down.
The feature is that it is performed using the representative value of the spectrum data of each unit.

With the above configuration, in addition to the operation of the configuration of claim 1, it is possible to perform scale-up and scale-down in more detail than 1 bit. Therefore, the calculation error can be reduced as compared with the case where scale-up and scale-down are performed in 1-bit units.

Since the representative value of the spectrum data of each unit is used, the scale-up and scale-down operations can be performed simultaneously with the inverse quantization process. Therefore, since the number of processing steps can be reduced as compared with the conventional case, it is possible to reduce the current consumption while removing the calculation error caused by the overflow.

In order to solve the above-mentioned problems, a digital data decoding apparatus according to a third aspect of the present invention, in order to solve the above-mentioned problems, a digital input signal is orthogonally transformed into spectrum data in the frequency domain for every unit time, and the spectrum data in the frequency domain in which the spectrum data is present. The spectrum data of each unit is quantized and encoded based on the representative value of the spectrum data of the unit to which it belongs, and then the quantized spectrum data and the representative value of the spectrum data of each unit In a digital data decoding device for decoding coded data recorded in a format including at least, the maximum value of the representative values of the spectrum data of each unit is searched, and the inverse quantum is calculated based on the maximum value. The optimal scaling is determined so that overflow does not occur in the process of conversion and inverse orthogonal transformation. It is characterized by scaling means for performing scaling up and scaling down is provided.

With the above configuration, it is possible to remove the calculation error caused by the overflow. In addition, in order to obtain the block floating scale amount, it is sufficient to search for the representative value of the spectrum data of each unit, which has a smaller number than the spectrum data. Since this representative value is an absolute value, the block floating scale amount can be calculated. The required processing steps can be shortened.

Therefore, since the number of processing steps can be reduced as compared with the conventional case, it is possible to reduce the current consumption while eliminating the calculation error caused by the overflow.

According to a fourth aspect of the present invention, in order to solve the above-mentioned problems, in addition to the configuration of the third aspect, the scaling means performs the above scale-up and scale-down for each unit. The feature is that the representative value of the spectrum data is used.

With the above configuration, in addition to the operation of the configuration of claim 3, scale-up and scale-down can be performed in more detail than 1 bit. Therefore, the calculation error can be reduced as compared with the case where scale-up and scale-down are performed in 1-bit units.

Since the representative value of the spectrum data of each unit is used, the scale-up and scale-down operations can be performed simultaneously with the inverse quantization process. Therefore, since the number of processing steps can be reduced as compared with the conventional case, it is possible to reduce the current consumption while removing the calculation error caused by the overflow.

[0023]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to FIGS. 1 to 9.

First, a minidisk recording / reproducing apparatus 1 equipped with a digital data decoding method and a decoding apparatus according to the present embodiment will be briefly described with reference to FIG. In the digital data decoding method and decoding apparatus of the present invention, a digital input signal is orthogonally converted into frequency domain spectrum data at every unit time, and the spectrum data is divided into frequency domain units. The spectrum data is quantized and encoded based on the representative value of the spectrum data of the unit to which it belongs, and then recorded in a format that includes at least the quantized spectrum data and the representative value of the spectrum data of each unit. The recording medium is not limited to a mini disk, and may be, for example, a digital video disk (DVD) because the encoded data is decoded.

As shown in FIG. 4, digital data is serially input to the input terminal 2 from a digital audio signal source such as a compact disk reproducing device or a satellite broadcast receiving device, for example, as an optical signal. The optical signal is converted into an electric signal by the photoelectric element 3 and then input to the digital PLL circuit 4. The digital PLL circuit 4 extracts a clock from the digital data and reproduces multi-bit data corresponding to the sampling frequency and the number of quantization bits. The multi-bit data corresponds to the mini disk standard in the frequency conversion circuit 5 from various sampling frequencies such as 44.1 kHz for a compact disk, 48 kHz for a digital audio tape recorder, or 32 kHz for satellite broadcasting (A mode). After the sampling rate is converted into the 44.1 kHz multi-bit data, it is input to the audio compression / decompression circuit 6.

The voice compression / decompression circuit 6 has a function as a digital data encoding device, and has a built-in ATR.
The AC processing circuit 6a compresses and encodes the input data according to the ATRAC method, and the encoded voice data is passed through the shock proof memory controller 7 to
It is input to the signal processing circuit 8. A shock proof memory 9 is provided in association with the shock proof memory controller 7. Shockproof memory 9
Absorbs the difference between the transfer rate of the audio data output from the audio compression / expansion circuit 6 and the transfer rate of the audio data input to the signal processing circuit 8, and reproduces a reproduction signal due to a disturbance such as vibration during reproduction described later. Is for interpolating the interruption of the and protecting the audio data.

The signal processing circuit 8 has a function as an encoder and a decoder, and encodes the audio data into a serial magnetic field modulation signal and supplies it to the head drive circuit 11. The head drive circuit 11 is the recording head 1.
2 is moved to a predetermined recording position on the mini disk 13 which is a magneto-optical disk, and a magnetic field corresponding to the magnetic field modulation signal is generated. At this time, the mini disk 13
The laser beam is emitted from the optical pickup 21 to the above-mentioned predetermined recording position, whereby a magnetization pattern corresponding to the magnetic field is formed on the mini disk 13.

On the other hand, from the mini disk 13, a serial signal corresponding to the magnetization pattern is reproduced by the optical pickup 21, the signal is amplified by a radio frequency (RF) amplifier 22, and then input to the signal processing circuit 8. Then, the audio data is decoded. The decoded audio data is sent to the shockproof memory controller 7 described above.
Then, after the influence of the disturbance is removed by the shock proof memory 9, it is inputted to the audio compression / expansion circuit 6. The audio compression / expansion circuit 6 has a function as a decoding device for digital data, and performs an inverse conversion process of compression encoding by the ATRAC system by the built-in ATRAC processing circuit 6a to demodulate into a full-bit digital audio signal. The demodulated digital audio signal is converted into an analog audio signal by the digital / analog (D / A) conversion circuit 23 and then output from the output terminal 24.

The serial signal amplified by the high frequency amplifier 22 is also input to the servo circuit 31, and the servo circuit 31 responds to the reproduced serial signal by way of the driver circuit 32. The rotation speed of 33 is feedback-controlled, which enables reproduction at a desired linear speed. Then, the rotation speed of the feed motor 34 is feedback-controlled, whereby the displacement of the optical pickup 21 in the radial direction of the mini disk 13, that is, tracking, can be controlled. Further, the focusing of the optical pickup 21 is feedback-controlled.

The servo circuit 31 and the optical pickup 2
1, the high frequency amplifier 22, the signal processing circuit 8, the driver circuit 32, and the like are energized by the power ON / OFF circuit 35. The system control microcomputer 36 centrally manages the power ON / OFF operation of the power ON / OFF circuit 35 and the signal processing operation described later.

Here, as described above, the mini disk recording / reproducing apparatus 1 has an ATR for audio compression / decompression.
The audio compression / decompression circuit 6 has a built-in ATRAC processing circuit 6a, which is a DSP that performs processing according to an AC (adaptive transform acoustic coding) method. This ATRAC
The method is a method used in a mini disk as a method for compressing and encoding digital data such as musical sounds and voices with high efficiency.

In the above ATRAC system, 44.1 kHz
The audio data sampled at is divided into predetermined frequency bands (low range of 0 to 5.5 kHz, middle range of 5.5 to 11 kHz, high range of 11 to 22 kHz), and then variable for each frequency band. MD is blocked by a long unit time
It is converted into spectrum data in the frequency domain by CT (modified discrete cosine transform) processing, and each spectrum data is encoded by the number of bits allocated by utilizing the psychoacoustic characteristics.

Next, a method of encoding digital data, which is a premise of the present invention and which encodes digital data according to the ATRAC system, will be briefly described with reference to FIGS. 5 and 7.

As shown in FIG. 7, in the mini-disc recording / reproducing apparatus 1 (FIG. 4), a band division filter section 101 and a time window cutting section 10 are provided as circuits for performing audio compression processing.
2, the MDCT unit 103, the bit allocation unit (WL calculation unit) 104, and the unit division unit (SF calculation unit) 105.
The quantization unit 106 and the encoding unit 107 are provided in the audio compression / expansion circuit 6.

First, the voice data sampled at 44.1 kHz is input to the voice compression / decompression circuit 6 and the QM
A frequency division filter unit 101 such as an F (quadrature mirror filter) divides the frequency band into two frequency bands (blocks) on the high frequency side and the low frequency side. The low-frequency side signal of the divided audio data further passes through the band division filter unit 101 again, and the frequency band is divided into two. As a result, the input voice data is eventually converted into the band division filter unit 1
By 01, it is divided into three bands of high band, middle band and low band (Fig. 5).

Next, the voice data divided into three bands is subjected to a maximum of 11.6 by the time window cutout unit 102.
It is cut out with a time window of ms (512 samples). Note that 512 samples of audio data are 424 bytes (2
× 212 bytes, 1 sound group).

Then, the MDCT unit 103 performs MDCT (modified discrete co
sine transform: Modified discrete cosine transform) processing is performed, and the sampled voice data is converted into spectrum data.

Subsequently, the bit allocation unit 104 performs bit allocation by applying psychoacoustic characteristics to the spectrum data, and the number of quantization bits (word length:
WL) is calculated. Further, in the unit dividing section 105, the spectrum data is further divided into a plurality of frequency regions (units). Here, the number of spectrum data included in each unit is different for each of the three frequency bands (blocks). Then, the amplitude magnitude (scale factor: SF) of the representative value (maximum value, etc.) of the spectrum data is calculated in units of the divided units.

FIG. 5 shows an example of how to take blocks and units. The low-frequency block is divided into 20 units by dividing it into four equal parts on the time axis and five equal parts on the frequency axis. Then, each unit of the low frequency block includes eight spectrum data. The mid-range block is divided into 16 units by equally dividing the time axis into four and the frequency axis into four. Then, each unit of the block in the middle range includes 6 pieces of spectrum data. The high-frequency block is divided into eight equal parts on the time axis and two equal parts on the frequency axis to be divided into 16 units. Then, each unit of the high frequency block includes 12 pieces of spectrum data. Therefore, 5
Twelve (0 to 511) spectrum data are divided into 52 (0 to 51) units. It should be noted that the way of taking the blocks and the units is an example, and the present invention is not limited to this.

Finally, the quantizing unit 106 quantizes the spectrum data with the word length and the scale factor, and then the coding unit 107 codes the compressed data.

The data encoded through the above procedure is
The data is recorded on the mini disk 13 in the compression format shown in FIG. In FIG. 6, one sound group (42
One channel of (4 bytes) is shown.

In the above compression format, BSM (block size m) including information on how to take blocks and units is used.
ode) and SIA (sub inform
ation amount) and word length index (WLI)
, Scale factor index (SFI), quantized spectrum (AS), sub scale factor index, sub word length index, sub SIA, and sub BSM from the beginning. They are stored in order. The contents of the sub information are almost duplicates of the main information.

Here, as shown in Table 1, the word length (W
L) is the number of quantization bits of each unit, and WL =
It is an integer of 0, 2 ≦ WL ≦ 16. Then, indexes (WLI) of 0 to 15 are added in order from the smallest,
This index is recorded.

[0044]

[Table 1]

As shown in Table 2, the scale factor (SF) is the magnitude of the amplitude of the representative value (maximum value etc.) of the spectrum data of each unit, and SF = m [i]
It is represented by × 2 ⁿ . However, m [1] = 0.62996052, m
[2] = 0.79370052, m [3] = 0.99999999, 1 ≦ i ≦
3, an integer of −5 ≦ n ≦ 16, and m when n = −5
[1] and m [2] are excluded. Then, indexes (SFI) of 0 to 63 are added in order from the smallest one, and this index is recorded. That is, the scale factor of each unit is scaled into 64 steps. Also, for one n, three coefficients m [1], m [2],
Since there are m [3], 1 bit corresponds to three parts of the scale factor index. The representative value of the spectrum data is, for example, the maximum value, and may be any other value as long as it replaces the maximum value.

[0046]

[Table 2]

8 and 9 show a method of decoding digital data, which is a premise of the present invention, for decoding encoded data recorded in the above-described compression format.
Will be briefly explained.

In the digital data decoding apparatus to which the digital data decoding method which is the premise of the present invention is applied, the WL expansion unit 111 is provided as a circuit for performing the voice expansion process.
, SF expansion section 112, AS expansion section 113, inverse quantization section 114, IMDCT section 115, windowing section 116
And a band synthesis filter unit 117 are provided in a circuit corresponding to the audio compression / decompression circuit 6. Further, the above I
The MDCT unit 115 includes a U (k) calculation unit 115a and an F (k) calculation unit 115a.
FT calculation unit 115b, u (n) calculation unit 115c, y
(N) A calculator 115d is provided.

First, the compressed data coded by the above coding process is input to a circuit corresponding to the audio compression / decompression circuit 6, and in the WL decompression unit 111, the word length of each unit in the block is calculated. Be expanded. Next, SF
In the expansion unit 112, the maximum value of the scale factor of each unit in the block is searched, and the IMDCT unit 1 after that is searched.
The scale amount of each unit is determined so that an overflow does not occur at 15. At this time, the scale amount is set according to the dynamic range accuracy of the scale factor. Further, the AS expansion unit 113 expands the quantized spectrum data into one word.

Subsequently, in the dequantization unit 114, dequantization is performed using the scale amount described above as an offset and a scale factor to which the offset has been added, and expanded into spectrum data.

Then, in the IMDCT unit 115, the spectrum data is stored in the U (k) calculation unit 115a and the FFT.
The calculation unit 115b, the u (n) calculation unit 115c, and the y (n) calculation unit 115d sequentially perform arithmetic processing, so that IMDCT (inverse modified discrete cosi) is performed in block units.
ne transform: Inverse transformation of modified discrete cosine) processing is performed and transformed into time axis data.

Finally, the windowing unit 116 adjusts the frames, and the band synthesizing filter unit 117 decodes it into voice data by a band synthesizing filter such as an IQMF (inverse quadrature mirror filter).

Next, decoding of digital data by the minidisk recording / reproducing apparatus 1 according to the present embodiment will be described. As shown in FIGS. 1 and 2, in the minidisk recording / reproducing apparatus 1, in particular, as a circuit for performing audio decompression processing, a WL expansion unit 41, an SF expansion unit 42, an AS expansion unit 43, and an SF maximum value determination. The audio compression / expansion circuit 6 is provided with a unit 44, an SF offset adding unit 45, an inverse quantization unit 46, an IMDCT unit 47, an inverse floating unit 48, a windowing unit 49, and a band synthesis filter unit 50. Has been. Further, the IMDCT unit 47 includes a U (k) calculating unit 47a, an FFT calculating unit 47b, a u (n) calculating unit 47c, a y (n) calculating unit 47d, a fixed scale down unit 47e, and a fixed scale down unit 47e. A scale-down value loading unit 47f is provided. The SF maximum value determination unit 44 described above
The SF offset applying section 45 and the inverse floating section 48 constitute a scaling means. Also,
SF maximum value determination unit 44 and SF offset attaching unit 45
Other than the IMDCT unit 47 and the inverse floating unit 48, the same configuration as the digital data decoding device which is the premise of the present invention shown in FIGS. 8 and 9 may be used.

Next, the digital data decoding process performed by the mini disk recording / reproducing apparatus 1 according to the present embodiment will be specifically described. Note that (2) to (6)
IMDCT processing shown in the definition formula of
DSP90-13, "A Study on MDCT Method and High-speed Arithmetic", Masahiro Iwadare, Takao Nishitani, Akihiko Sugiyama, et al., And I
EEE, Transactions on ASSP, Vol. 34, No. 5, "Analysis / Synthesis Filterbank Design Based on Aliasing Cancellation in Time Domain", John P. Princen, Alan Bernard B
See radley for details.

First, when the encoded data recorded in the above-mentioned compression format is input to the ATRAC processing circuit 6a, the WL expansion section 41 determines the word length index of each unit stored as shown in FIG. , Is expanded (replaced) into a numerical value of the actual word length according to Table 1 and stored in the memory.

Similarly, in the SF expansion section 42, the scale factor index of each unit stored as shown in FIG. 6 is expanded (replaced) into an actual scale factor value according to Table 2 and stored in the memory. To do.

Next, the AS expansion unit 43 expands the quantized spectrum data stored as shown in FIG. 6 for each word (16 bits). here,
The AS expansion unit 43 reads out the stored quantized spectrum data in units of 8 bits. The maximum word length is 16 bits. Therefore, 1 word is 1
There are cases in which reading can be performed by reading once, reading by reading twice, and reading by reading three times.
That is, the processing content of the AS expansion unit 43 is to read out the quantized spectrum data stored in various states by 8 bits and cut it by 1 word (maximum 16 bits).

Next, in the SF maximum value determining section 44,
The maximum value of the included scale factor is searched for for each block. Here, the block is a high frequency band divided into three bands by the band division filter unit 101,
It refers to the mid-range and low-range frequency bands. And later I
To prevent overflow in MDCT unit 47,
The block scale is optimally determined. At this time,
The scale amount is set by the dynamic range accuracy of the scale factor.

Here, there are a maximum of 52 scale factors within a block because they exist for each unit included. On the other hand, the maximum number of quantized spectrum data is 512 because the number of quantized spectrum data exists. The unit also contains at least one piece of quantized spectrum data. Therefore, when detecting the maximum value of the scale factor, it is clear that the number of scale factors in the block is smaller than the number of quantized spectrum data.

In the SF offset attaching unit 45, the maximum value of the scale factor included in each block is SF = m [3].
The scale amount is calculated so as to be × 2 ¹⁶ , and this is used as the offset value of the scale factor index. Then, the obtained offset values are added to the indexes of all scale factors in the block, and the indexes of the scale factors are updated. The index of this offset value and the updated scale factor is A
It is stored in a memory or the like in the TRAC processing circuit 6a and used in the reverse floating unit 48 later.

Here, a specific description will be given using Table 2.
The maximum value of the scale factor in the block is SF = m
Consider the case of [1] × 2 ¹⁰ = 0.62996052 × 2 ¹⁰ . The maximum value of this scale factor has an index of 43. Then, the maximum value of this scale factor is SF =
When the offset value of the index is calculated so that m [3] × 2 ¹⁶ = 0.99999999 × 2 ¹⁶ , the offset value =
63-43 = 20.

Now, a certain scale factor SF = m [2] × 2 ⁶ = 0. Existing in the same block as the above maximum value.
Consider 79370052 × 2 ⁶ (SFI = 32). When the above offset value is added to this index and updated, the index is SFI = 32 + 20 = 52 (SF = m
[1] × 2 ¹³ = 0.62996052 × 2 ¹³ ). Similarly, update the index of all scale factors in the block.

Next, in the inverse quantizer 46, the W
The inverse quantization process of Expression (1) is performed by the word length obtained by the L expansion unit 41 and the scale factor indicated by the index of the scale factor updated by the SF offset assigning unit 45. As a result, the quantized spectrum data obtained by the AS expansion unit 43 is inversely quantized into spectrum data.

[0064]

[Equation 1]

Here, i is the unit number and l is the spectrum data number. Note that i <l.

Subsequently, in the IMDCT unit 47, the spectrum data dequantized by the dequantization unit 46 is
IMDCT processing is performed according to the equations (2) to (6) in units of blocks in the high frequency band, the middle frequency band, and the low frequency band, and is converted into time axis data. It should be noted that FIG. 2 summarizes the processing of the expressions (2) to (6).

[0067]

[Equation 2]

[0068]

[Equation 3]

However, Re [Z (l)], Im [Z
(L)] indicates the real and imaginary parts of the complex number Z [l].

[0070]

[Equation 4]

However, i represents an imaginary unit. In addition,
A log ₂ (M / 2) stage butterfly operation is often used for the operation processing of the equation (4).

[0072]

[Equation 5]

[0073]

[Equation 6]

Here, before performing the respective arithmetic processing of the IMDCT conversion, the fixed scale down unit 47e
The data to be operated is scaled down according to the fixed scale down value table prepared in advance in the audio compression / decompression circuit 6. That is, before performing each calculation process of the IMDCT conversion, the fixed scale down value loading unit 47f sets the fixed scale down value shown in Table 5 (described later) according to the maximum value of the scale factor in each block.
The data is loaded into a memory or the like in the ATRAC processing circuit 6a, and the fixed scale-down unit 47e scales down the operated data by the fixed scale-down value.

A detailed description will be given according to the above example. The maximum value of the scale factor in the block is SF = m [1]
× Being an 2 ^10, the maximum value of the scale factor in Table 5 are with reference to the column of 2 ^10, (3) before treatment to Expression (5), after a fixed scale down, IMDCT Each calculation process of conversion is performed. That is, 1 bit is shifted to the right before the equation (3), 1 bit is shifted to the right before the FFT first stage, FFT second stage, and FFT third stage of the formula (4), and right is shifted before the FFT fourth stage. Shift 2 bits to (5)
Do not shift before the expression.

In this embodiment, the fixed scale down value shown in Table 5 is used, but the fixed scale down value used here may be any value which does not cause an overflow in the process of each arithmetic processing of IMDCT conversion. , But not limited to those in Table 5.

Now, Table 5 referred to by the fixed scale down value loading unit 47f will be described. First,
In equations (2) to (6),
Table 3 shows how many times the data after calculation increases. The number of spare bits in Table 3 is determined by design. Since this is known from Japanese Patent Application Laid-Open No. 7-36666, only the result is shown and the details are omitted.

[0078]

[Table 3]

Table 4 shows the ratio of the operated data to the operated data. The ratios in Table 4 are determined by design. Incidentally, this is also disclosed in JP-A-7-366.
Since it is known from Japanese Patent No. 66, only the result is shown
Details are omitted.

[0080]

[Table 4]

Then, in Table 3, Table 4, and the following three conditions, Table 5 shows the size of the data to be operated and the optimum fixed scale-down amount corresponding to each operation processing. The maximum input value of IMDCT is 2 ¹⁶ . The maximum output value of IMDCT is 2 ¹⁵ . From the above two conditions and each magnification and each ratio, IMD
There is a maximum data value in each calculation process of CT conversion. This is obtained by drawing an envelope curve representing the maximum value of the input data of the IMDCT conversion, that is, the maximum value in each processing step when the maximum value of the scale factor is 2 ⁿ . The FFT is an example of four stages. Incidentally, this is also disclosed in JP-A-7-
Since it is known from Japanese Patent No. 36666, only the result is shown and the details are omitted.

[0082]

[Table 5]

Next, in the inverse floating unit 48, after the inverse orthogonal transform process in the IMDCT unit 47, the scale factor is SF = m [3] × 2 ¹⁶ , that is, SFI =
Return the scale amount (offset) based on 63,
End block floating. That is, the above I
The MDCT unit 47 subtracts a value obtained by multiplying the total sum of the fixed scale-down performed before the processing of Expressions (3) to (5) by 3 from the offset value of the scale factor index, and updates the scale factor offset value. . Calculate the inverse scale amount from this updated offset value,
The inverse floating is terminated by multiplying the data after the IMDCT conversion.

Here, the scale factor is 1 dB to 3
It is set at equal intervals. That is, the unit of 1 bit corresponds to 3 ranges in the scale factor index (1 bit = 3 indexes). As a result, the value obtained by multiplying the total sum of the fixed scale-down bit numbers by 3 is subtracted from the offset value of the scale factor index.

A detailed description will be given according to the above example. First, the total sum of fixed scale down is calculated. Since the maximum value of the scale factor in the block is at SF = m [1] × 2 10, the maximum value of the scale factor in Table 5 refers to the column of 2 ^10, the IMDCT unit 47 (3) -
The total number of fixed scale-downs performed before the processing of equation (5) is added in order from the top, giving 1 + 1 + 1 + 1 + 2 = 6. When this is multiplied by 3, it becomes 18. Therefore, the scale factor index may be lowered by 18. That is,
Since the stored offset value is 20, the offset value is updated to 20− (6 × 3) = 2. further,
When the inverse scale amount is calculated based on SFI = 63, SFI = 63-2 = 61. Therefore, the inverse floating is terminated by multiplying the scale factor SF = m [1] × 2 ¹⁶ indicated by the scale factor index with the data after the IMDCT conversion. Note that the multiplication is fixed point.

Further, the windowing section 49 performs windowing processing on the IMDCT processed data one sample before on the time axis. As a result, adjustment between adjacent frames on the time axis is performed.

Finally, in the band synthesizing filter unit 50, the IQMF which is a band synthesizing filter decodes the audio data, and the decoding process of the digital data ends.

The above-mentioned IMDCT section 47 (FIG. 1)
Can also be configured as shown in FIG. That is,
The IMDCT unit 47 includes a U (k) calculation unit 47a ′ and an F (k) calculation unit 47a ′.
FT calculator 47b ', u (n) calculator 47c', y
(N) Calculation unit 47d 'and scale-down unit 47e'
And a scale-down value determination unit 47f 'are provided. In the IMDCT unit 47, the inverse quantization unit 46
Inverse quantized spectrum data in high range, mid range,
Similar to the configuration shown in FIG. 2, the IMDCT process is performed in the low frequency block unit according to the equations (2) to (6) and converted into time axis data. Therefore, in comparison with the configuration shown in FIG. 2, a scale down unit 47e ′ is provided instead of the fixed scale down unit 47e, and a scale down value determination unit 47f ′ is provided instead of the fixed scale down value loading unit 47f. It may have the same configuration. Further, this configuration is disclosed in Japanese Patent Laid-Open No. 6-164.
Since it is known from Japanese Patent Publication No. 414, the details will be omitted.

In the IMDCT unit 47 having the configuration shown in FIG. 3, the scale-down value determination unit 47f 'obtains the maximum absolute value of the operated data of each block before performing each operation process of the above IMDCT conversion, The scale-down value is determined so that the obtained maximum absolute value does not overflow even when multiplied by the scaling factor (Table 3) prepared in advance in the audio compression / decompression circuit 6. And the scale down section 4
In 7e ', the data to be calculated is scaled down or scaled up by the obtained scale down value.

Furthermore, when the IMDCT conversion is performed while the variable scale down is performed as described above, the inverse floating unit 48 calculates the offset value of the index of the scale factor before the calculation processing of the expressions (3) to (5). The value obtained by multiplying the total sum of the scale-down values performed by 3 is subtracted, and SF
Update the offset value of I. The other processing is the same as that described above.

Here, Table 5 referred to by the fixed scale-down unit 47f is fixed with a value that does not cause an overflow even in the worst case. Further, in Table 5, when the maximum scale factor value is 2 ⁹ or less, the scale down values are exactly the same. In other words, small signals cannot be scaled up any further, resulting in a very weak scale. For the above reasons, by configuring the IMDCT unit 47 to perform variable scale down, good calculation accuracy can be obtained.

For the same reason, the IMDCT unit 47
Can be configured to switch the processing of determining the scale down value according to the maximum value of the scale factor. That is, when the maximum scale factor is 2 ¹⁰ or more, fixed scale down is performed, and when the maximum scale factor is 2 ⁹ or less, variable scale down is performed. As a result, better calculation accuracy can be obtained.

As described above, according to the audio compression / expansion circuit 6 provided in the mini-disc recording / reproducing apparatus 1 according to the present embodiment, based on the maximum value of the scale factor of each unit, for each frequency band. The scale of block floating is optimally determined so that overflow does not occur in the process of inverse quantization, inverse orthogonal transform, and scale up and scale down are performed.

As a result, the calculation error caused by the overflow can be removed. Further, in order to obtain the block floating scale amount, it is sufficient to search for a scale factor that is smaller in number than the spectrum data, and since the scale factor is an absolute value, the processing step for obtaining the block floating scale amount can be shortened.

Therefore, since the number of processing steps can be reduced as compared with the conventional case, it is possible to reduce the current consumption while eliminating the calculation error caused by the overflow.

Further, according to the audio compression / expansion circuit 6 provided in the mini-disc recording / reproducing apparatus 1 according to the present embodiment, the above scale-up and scale-down are performed using the scale factor.

As a result, more detailed scale-up and scale-down than 1 bit can be performed. Therefore,
The calculation error can be reduced as compared with the case where scale-up and scale-down are performed in 1-bit units.

Since the scale factor is used, the scale-up and scale-down operations can be performed simultaneously with the inverse quantization process. Therefore, since the number of processing steps can be reduced as compared with the conventional case, it is possible to reduce the current consumption while removing the calculation error caused by the overflow.

[0099]

As described above, the method for decoding digital data according to the first aspect of the present invention is such that the digital input signal is orthogonally transformed into spectrum data in the frequency domain for each unit time, and the spectrum data in the frequency domain in which the spectrum data exists is obtained. The spectrum data of each unit is quantized and encoded based on the representative value of the spectrum data of the unit to which it belongs, and then the quantized spectrum data and the representative value of the spectrum data of each unit In a digital data decoding method for decoding coded data recorded in a format having at least, the maximum value of the representative values of the spectrum data of each unit is searched, and the inverse quantum is calculated based on the maximum value. The scale amount is optimally determined so that overflow does not occur in the process of conversion It is configured to perform pull-up and scale-down.

Therefore, the calculation error caused by the overflow can be removed. In addition, in order to obtain the block floating scale amount, it is sufficient to search for the representative value of the spectrum data of each unit, which has a smaller number than the spectrum data. Since this representative value is an absolute value, the block floating scale amount can be calculated. The required processing steps can be shortened.

Therefore, since the number of processing steps can be reduced as compared with the conventional case, there is an effect that it is possible to reduce the current consumption while eliminating the calculation error caused by the overflow.

As described above, the digital data decoding method according to the second aspect of the present invention uses the representative value of the spectrum data of each unit for the above scale-up and scale-down in addition to the configuration of the first aspect. This is a configuration to be performed.

Therefore, in addition to the effect of the configuration of claim 1, scale-up and scale-down can be performed in more detail than 1 bit. Therefore, rather than scaling up and scaling down in 1-bit units,
It is possible to reduce the calculation error.

Since the representative value of the spectrum data of each unit is used, the scale-up and scale-down operations can be performed simultaneously with the inverse quantization process. Therefore, since the number of processing steps can be reduced as compared with the conventional technique, it is possible to reduce the consumption current while removing the calculation error caused by the overflow.

As described above, in the digital data decoding device of the third aspect of the present invention, the digital input signal is orthogonally transformed into frequency domain spectrum data at every unit time, and the spectrum data is converted into a frequency domain unit. Based on the representative value of the spectrum data of the unit to which the spectrum data of each unit is divided,
After being quantized and coded, the quantized spectrum data, and a digital data decoding device for decoding the coded data recorded in a format having at least a representative value of the spectrum data of each unit, The maximum value of the representative values of the spectrum data of each unit described above is searched, and the scale amount is optimally determined and scaled up based on the maximum value so that overflow does not occur in the process of inverse quantization and inverse orthogonal transform. And a scaling means for performing scale down.

Therefore, the calculation error caused by the overflow can be removed. In addition, in order to obtain the block floating scale amount, it is sufficient to search for the representative value of the spectrum data of each unit, which has a smaller number than the spectrum data. Since this representative value is an absolute value, the block floating scale amount can be calculated. The required processing steps can be shortened.

Therefore, since the number of processing steps can be reduced as compared with the conventional case, the effect that the current consumption can be reduced while eliminating the calculation error caused by the overflow is obtained.

As described above, in the digital data decoding device according to the fourth aspect of the present invention, in addition to the configuration of the third aspect, the scaling means performs the scale-up and the scale-down by the spectrum data of each unit. The configuration is performed using the representative value of

Therefore, in addition to the effect of the structure of claim 3, scale-up and scale-down can be performed in more detail than 1 bit. Therefore, rather than scaling up and scaling down in 1-bit units,
It is possible to reduce the calculation error.

Since the representative value of the spectrum data of each unit is used, the scale-up and scale-down operations can be performed simultaneously with the inverse quantization process. Therefore, since the number of processing steps can be reduced as compared with the conventional technique, it is possible to reduce the consumption current while removing the calculation error caused by the overflow.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing an outline of a circuit configuration for performing a decoding process of digital data provided in a voice compression / decompression circuit according to an embodiment of the present invention.

FIG. 2 is a block diagram of I provided in the audio compression / decompression circuit shown in FIG.
It is a block circuit diagram showing a configuration of an MDCT unit.

FIG. 3 shows I provided in the audio compression / expansion circuit shown in FIG.
FIG. 9 is a block circuit diagram showing another configuration of the MDCT unit.

FIG. 4 is a block diagram showing an outline of a configuration of a mini disk recording / reproducing apparatus including an audio compression / expansion circuit according to an embodiment of the present invention.

FIG. 5 is an explanatory diagram showing an example of how blocks and units of modified discrete cosine transform (MDCT) are captured.

FIG. 6 is an explanatory diagram showing an outline of a main part of a compression format of a mini disc.

FIG. 7 is an explanatory diagram showing an outline of a circuit configuration of a digital data encoding device which is a premise of the present invention.

FIG. 8 is an explanatory diagram showing an outline of a circuit configuration of a digital data decoding device which is a premise of the present invention.

9 is a block circuit diagram showing a configuration of an IMDCT unit provided in the digital data decoding apparatus which is the premise of the present invention shown in FIG.

[Explanation of symbols]

1 Mini-disc recording / reproducing device 6 Audio compression / expansion circuit (digital data decoding device) 6a ATRAC processing circuit 41 WL expansion part 42 SF expansion part 43 AS expansion part 44 SF maximum value determination part (scaling means) 45 SF offsetting part (Scaling means) 46 Inverse quantization section 47 IMDCT section 48 Inverse floating section (scaling means) 49 Windowing section 50 Band synthesis filter section 47a U (k) calculation section 47b FFT calculation section 47c u (n) calculation section 47d y ( n) calculation unit 47e fixed scale down unit 47f fixed scale down value loading unit 47a 'U (k) calculation unit 47b' FFT calculation unit 47c 'u (n) calculation unit 47d' y (n) calculation unit 47e 'scale down unit 47f 'Scale-down value determination unit

─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP-A-6-164414 (JP, A) JP-A-6-318875 (JP, A) JP-A-7-36666 (JP, A) JP-A-5- 55925 (JP, A) JP 63-136826 (JP, A) JP 63-152229 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) H03M 7/30 G06F 7 / 38 G11B 20/10 301

Claims (4)

(57) [Claims] 1. A digital input signal is orthogonally transformed into frequency domain spectrum data for each unit time, and the spectrum data is divided into frequency domain units.
A format in which the spectrum data of each unit is quantized and encoded based on the representative value of the spectrum data of the unit to which it belongs, and at least the quantized spectrum data and the representative value of the spectrum data of each unit. In the digital data decoding method for decoding the encoded data recorded in, the maximum value of the representative value of the spectrum data of each unit is searched, and the inverse quantization and the inverse orthogonal transform are performed based on the maximum value. A method of decoding digital data, characterized in that the scale amount is optimally determined so that overflow does not occur in the process of step 1, and scale-up and scale-down are performed. 2. The method of decoding digital data according to claim 1, wherein the scale-up and scale-down are performed by using a representative value of spectrum data of each unit. 3. A digital input signal is orthogonally transformed into frequency domain spectrum data at every unit time, and the spectrum data is divided into frequency domain units.
A format in which the spectrum data of each unit is quantized and encoded based on the representative value of the spectrum data of the unit to which it belongs, and at least the quantized spectrum data and the representative value of the spectrum data of each unit. In a digital data decoding device that decodes the encoded data recorded in, the maximum value of the representative value of the spectrum data of each unit is searched, and the inverse quantization and the inverse orthogonal transform are performed based on the maximum value. An apparatus for decoding digital data, characterized in that it is provided with scaling means for optimally determining a scale amount so that overflow does not occur in the process of (1) and performing scale-up and scale-down. 4. The digital data decoding apparatus according to claim 3, wherein said scaling means performs said scale-up and scale-down by using a representative value of spectrum data of each unit.