JP3064471B2 - Orthogonal transform coding method and inverse orthogonal transform 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 an orthogonal transform coding method for extracting and orthogonally transforming an input signal by applying an analysis window, and performing inverse orthogonal transform on an orthogonally transformed signal, applying a synthesis window, and connecting to an adjacent block. And to an inverse orthogonal transform decoding device for restoring an original waveform.

[0002]

2. Description of the Related Art Briefly describing orthogonal transform coding, a digital signal obtained by digitizing an audio signal, a video signal, or the like is converted into, for example, a discrete Fourier transform (DF) for each predetermined number of samples (blocks or frames).
T) or the like, and quantizes the obtained coefficient data on the frequency axis.

Here, in the above-mentioned block formation, an analysis window (window) AW is applied so that an overlap portion (overlap portion) OL occurs between adjacent blocks as shown in FIG. Windowing (Win
dowing). In FIG. 8, as a specific example of the analysis window AW, a window function having a so-called trapezoidal shape including a linear portion a having a value of 1.0 and inclined portions b at both ends is used. One block (or one frame) extends from one end where the value of the window function is 0.0 to the other end where the value becomes 0.0 again through the inclined portion b, the straight line portion a, and the inclined portion b. The sample data in one block BL is subjected to the orthogonal transform processing such as the DFT.

[0004] The reason for using such an analysis window AW is that when an adjacent block is divided without overlapping and orthogonally transformed,
This is because a phenomenon in which the signal at the end of the block after the inverse orthogonal transform is wrapped around, that is, a so-called end effect occurs, and adversely affects the sense of hearing or the sense of sight. Therefore, generally, the orthogonal transform encoding process is performed in a procedure as shown in FIG. 9A, and the inverse orthogonal transform decoding process is performed in a procedure as shown in FIG. 9B.

That is, on the encoder side shown in FIG. 9A, the waveform extracting section 111 extracts data from the input waveform data for each block including the overlapping portion OL, and the next analysis windowing section 112 extracts one block. The above analysis window AW is applied to the minute data. Next, orthogonal transform such as DFT is performed by the orthogonal transform unit 113, and encoding such as quantization is performed by the encoding unit 114, so that encoded data on the frequency axis is obtained. Applying an analysis window before performing the orthogonal transformation (so-called analysis
Windowing) prevents the spectrum from spreading.

[0009] Next, in FIG. 9 B, the coded data on the frequency axis is decoded by a decoding unit 121, subjected to inverse orthogonal transformation by an inverse orthogonal transformation unit 122, and then to a synthesis windowing unit 123. The synthesis window SW is set at
ndowing). The synthesis window SW has the same shape as the analysis window AW, and is for suppressing noise at both ends of the block. The signal of each block to which the synthesis window SW has been applied is sent to the waveform connection unit 124, and the above overlapped portions are added and connected to restore the original waveform.

Here, a specific example of each function of the analysis window AW and the synthesis window SW will be described with reference to FIG. The analysis window AW of FIG. 10 is a function w of the following equation (1).
₁ (n), and the synthesis window SW uses the function w ₂ (n) of the following equation (2). Further, a function w (n) obtained by multiplying the analysis window function w ₁ (n) and the composite window function w ₂ (n) is shown in the following equation (3), and FIG. 10 shows the multiplied waveform ASW.

(Equation 1)

[0008]

By the way, conventionally, as is apparent from the above equations (1) and (2),
The same function is used for the analysis window function w ₁ (n) and the composite window function w ₂ (n). In the functions shown in the equations (1) and (2), the non-overlapping portion NOL takes a value of 1.0,
The overlap portion OL at both ends is obtained by taking the square root of each of the two portions divided left and right around a so-called Hanning Window function, and a function w (n) shown in the equation (3) obtained by multiplying the two is obtained. , Overlapped portions OL at both ends are left and right portions of the above-described two-part Hanning Window function.

However, in such a conventional windowing, the Hanning Window function is used as the analysis window function w ₁ (n).
The shoulder of the function waveform, which takes the square root of the function and takes a value other than 1.0, is the overlapped portion O at the both ends.
Since only L is short, it cannot be sufficiently smoothly converged to 0 at both ends. For this reason, the spectrum of the output that has been orthogonally transformed using such an analysis window is spread on the frequency axis, and many bits are required to accurately encode these spectrum output data. There are drawbacks.

Also, considering that the shoulder portion is short, if the overlap portion OL is lengthened to increase the shoulder portion, the spectrum of the orthogonal transform output for each block is concentrated on the frequency axis. However, the number of blocks per fixed time increases due to an increase in the overlapping portion between adjacent blocks on the time axis, and the bit rate required to secure a predetermined S / N increases, which is preferable. Absent. Conversely, when the overlap portion OL is shortened, the spectrum on the frequency axis is widened, and the number of bits required for encoding with predetermined accuracy increases.

The present invention has been proposed in view of the above-described circumstances, and has a smoother shoulder portion of an analysis window without increasing the width of an overlapping portion with an adjacent block. Focus the spectrum on the frequency axis,
It is an object of the present invention to provide an orthogonal transform coding method and an inverse orthogonal transform decoding device that can enhance the bit reduction effect without giving any adverse effect.

[0012]

According to the orthogonal transform coding method of the present invention, an input signal is divided into blocks on a time axis by applying an analysis window, and each block is subjected to orthogonal transform and encoded. In the orthogonal transform coding method, as the analysis window, a window function having a shoulder having a width wider than an overlapping portion with a block adjacent on the time axis is used at both ends, and on the frequency axis after the orthogonal transform. The above-mentioned problem is solved by concentrating the spectrum at.

Further, the inverse orthogonal transform decoding apparatus according to the present invention provides an orthogonal transform code after applying an analysis window having a shoulder wider at both ends than a portion overlapping with a block adjacent on the time axis. The signal on the converted frequency axis is input, the input signal is subjected to inverse orthogonal transform to be converted into a signal on the time axis, and the result of multiplication with the window function of the analysis window is used as the overlapped portion of the adjacent block. A composite window of a window function that becomes 1 when added is applied to the signal subjected to the inverse orthogonal transform, and the signal applied with the composite window is connected to a signal of an adjacent block to form an original signal. The above-mentioned problem is solved by restoring the signal on the time axis.

[0014]

Since the analysis window and the synthesis window are different, and the analysis window is windowed by using a window having a shoulder wider than an overlapping portion of an adjacent block, the block is formed by windowing. In addition, the spectrum after the orthogonal transform can be more intensively distributed on the frequency axis, and accurate coding can be realized with a small number of bits. In this case, since the width of the overlapping portion is not increased, the redundancy due to the overlapping between adjacent blocks does not increase, and the bit rate does not increase.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an analysis window AW for windowing an input signal on the frequency axis at the time of encoding (orthogonal transform coding) and an inverse orthogonal at the time of decoding (inverse orthogonal transform decoding). A synthesis window SW for windowing the signal on the frequency axis obtained by the conversion,
A multiplied waveform ASW obtained by multiplying each of the window functions of the analysis window AW and the synthesis window SW is one block B
L is shown.

As is apparent from FIG. 1, the analysis window AW
Has a shoulder SH wider than an overlapping portion OL with an adjacent block (that is, both ends having a value smaller than 1 and larger than 0). The composite window SW for the analysis window AW becomes 1 when the overlap portion OL of the multiplication waveform ASW is added to the overlap portion OL of the adjacent block as a waveform ASW of a function obtained by multiplying these window functions. It has become.

FIG. 2 shows an example of a configuration for encoding (orthogonal transform coding). A signal on a time axis supplied to an input terminal 11, for example, waveform data such as an audio PCM signal is temporarily stored in a memory 12. The overlap OL portion is read out so as to overlap the data of the previous block for each block BL, and is subjected to a so-called waveform cutout into a block. The data for each block BL read from the memory 12 is sent to the analysis windowing circuit 13 so that the analysis window generation circuit 14
The above analysis window AW is applied. The data to which the analysis window AW has been applied is sent to an orthogonal transform circuit, for example, an FFT (fast Fourier transform) circuit 15, and subjected to FFT processing for each block, thereby becoming spectrum data on the frequency axis. The orthogonally transformed output data on the frequency axis is sent to the encoder 16 where it is encoded and extracted via the output terminal 17.

FIG. 3 shows a configuration example for decoding (inverse orthogonal transform decoding). 3, the input terminal 21 is supplied with coded data orthogonally coded by the decoding configuration shown in FIG. 2 and sent to a decoder 22 to be decoded. This is so-called FFT coefficient data. This data is converted into an inverse orthogonal transform circuit, for example, IFFT (Inverse Fast Fourier Transform)
The data is sent to the circuit 23 and subjected to the reverse processing of the FFT processing, that is, the conversion processing from the frequency axis to the time axis, and becomes data on the time axis in block units. The data in block units is sent to the composition windowing circuit 24, and the composition window generation circuit 25 applies the composition window SW. The output from the synthesis windowing circuit 24 is obtained by multiplying the original waveform data by the multiplied waveform ASW, and is sent to the next waveform connection circuit 26 to be superimposed on the adjacent block. By adding and connecting the (lap) portions OL, continuous waveform data on the original time axis is restored.

A specific example of the window function w ₁ (n) of the analysis window AW is shown in the following equation (4), and a specific example of the function w ₂ (n) of the synthesis window SW is shown in the following equation (5). Is shown. A function w (n) obtained by multiplying the analysis window function w ₁ (n) and the composite window function w ₂ (n) is shown in the following equation (6).

(Equation 2)

In these equations (4) to (6), N represents the length of one block BL by the number of samples, and the horizontal axis (time axis) in FIG. When the leading end (left end in the figure) is 0, block B
The end of L (the right end in the figure) is N. M ₁ represents the width of the overlapping portion OL with the adjacent block (overlap) by the number of samples, and M ₂ represents the shoulder of the analysis window (that is, the value between 0 and 1 at both ends of the function shape). (Part to be taken) The width of SH is represented by the number of samples. Then, the width M ₂ of the shoulder portion SH is widely (M _2> M ₁₎ setting than the width M ₁ of the overlap portion OL, in the specific example of FIG. 1, M
₂ = and has a 1.5M _1. Also, on the time axis (sample number) of the window function w ₁ (n) in equation (4), 0 to M ₂ , NM
₂ the parts of to N (corresponding to the shoulder portion SH of the left and right ends of the analysis window AW of Fig. 1) is a so-called Hanning window (Ha width of 2M ₂
n (nning Window) function corresponds to the left and right parts obtained by dividing the function into two in the center, and 0 to M ₁ of the multiplication function w (n) in the equation (6).
Each of the portions NM _{1 to} N (each overlap portion OL of the multiplied waveform ASW in FIG. 1) corresponds to each of left and right portions obtained by dividing a Hanning window function having a width of 2M ₁ into two at the center.

FIG. 4 shows an input signal, for example, 1 k
FIG. 5 shows a spectrum distribution on the frequency axis obtained by applying the analysis window AW of FIG. 1 to the sine wave signal of Hz and then orthogonally transforming the sine wave signal. FIG. 12 shows a spectrum distribution on a frequency axis obtained by applying an analysis window AW (of FIG. 10) and then performing orthogonal transformation. As is apparent from FIGS. 4 and 5, when the windowing process is performed using the analysis window AW shown in FIG. 1 of the embodiment of the present invention, the spectral data Are concentrated on the frequency axis. Therefore, when such spectrum data is subjected to encoding processing including requantization processing and the like and transmitted, highly accurate encoding processing can be performed with a small number of bits. In other words, in the case of the spectrum data as shown in FIG. 5 in which the conventional windowing process is performed, since the spectrum is spread on the frequency axis, it is necessary to encode each spectrum data with high accuracy. In the example of FIG. 4 in which the windowing process according to the embodiment of the present invention is performed while the number of bits to be allocated increases, the number of bits to be allocated to obtain the same accuracy can be reduced because the degree of spectral concentration is high. is there. In addition, since the overlap width between adjacent blocks is the same as in the related art, there is no increase in redundancy due to block overlap on the time axis, and an increase in bit rate can be avoided.

Next, FIG. 6 shows a specific example of a high-efficiency coding apparatus which mixes band division coding and adaptive conversion coding as an orthogonal transform coding apparatus to which the present invention can be applied.
In FIG. 6, for example, 0 to 20 k
Hz audio PCM signal is supplied. This input signal is divided into a band of 0 to 10 kHz and a band of 10 to 20 kHz by a band division filter 31 such as a so-called QMF filter.
And the signal in the 0 to 10 kHz band is also Q
0-5k by band division filter 32 such as MF filter
It is divided into a Hz band and a 5 kHz to 10 kHz band. The signal in the 10 kHz to 20 kHz band from the band division filter 31 is sent to a fast Fourier transform (FFT) circuit 33 which is an orthogonal transformation circuit, and the signal in the 5 kHz to 10 kHz band from the band division filter 32 is sent to an FFT circuit 34. The signals in the 0 to 5 kHz band from the band division filter 32 are sent to the FFT circuit 35 to be subjected to FFT processing.

FIG. 7 shows a specific example of blocking for each band supplied to each of the FFT circuits 33, 34 and 35.
In the specific example of FIG. 7, the frequency band is widened and the time resolution is increased (the block length is shortened) toward the higher frequency side.
ing. That is, for a signal in the 0 to 5 kHz band on the low frequency side, one block BL _L1 is, for example, 1024 samples, overlap portions OL are provided at both ends, and a non-overlap portion NOL is provided therebetween.
Further, the signal in the 5 kHz to 10 kHz band in the middle band is divided into blocks BL _M1 and BL _M2 each having a half length of the block BL _{L1 in} the low band, and the blocks 10 _M to 20 _M in the high band are used.
For signals in the kHz band, the low-frequency block BL _L1
Of blocks BL _H1 , BL _H2 , B each having a length of １ ／
Blocked by L _H3 and BL _H4 . The actual shape of the analysis window at the time of blocking is, of course, a shoulder portion longer than the overlap portion OL in FIG. When a band of 0 to 22 kHz is considered as an input signal, the low band is 0 to 5.5 kHz and the middle band is 5.5 kHz.
１１11 kHz, and the high frequency range is 11 to 22 kHz.

Referring again to FIG. 6, each of the FFT circuits 33, 3
The spectrum data or FFT coefficient data on the frequency axis subjected to the FFT processing from Nos. 4 and 35 are collected for each so-called critical band (critical band) and sent to the bit adaptive allocation encoding circuit 36. The critical band is a frequency band divided in consideration of human auditory characteristics, and a band of a pure tone when the pure tone is masked by a narrow band noise near the frequency of the pure tone. That is. In this critical band, the higher the band, the wider the bandwidth.
The entire frequency band of ２０20 kHz is divided into, for example, 25 critical bands.

The allowable noise calculating circuit 37 calculates an allowable noise amount for each critical band in consideration of a so-called masking effect and the like based on the spectrum data (FFT coefficient data) divided into the critical band. And the number of bits to be allocated for each critical band based on the energy or the peak value for each critical band, and the like.
, Each spectrum data (FFT coefficient data) according to the number of bits allocated to each critical band
Is requantized. The data encoded in this manner is taken out via the output terminal 38.

The present invention is not limited to the above embodiment. For example, the analysis window function and the like are not limited to the Hanning window function, but may be trapezoidal or other various window functions. One with a longer (wider) shoulder than the overlap may be used. In addition, the width of the shoulder of the analysis window can be increased at various magnifications other than 1.5 times the width of the overlap portion. However, when the magnification is increased, the value of the window function of the synthesis window becomes large at both ends, and noise at the time of decoding increases. Therefore, it is preferable to limit the window function to, for example, about 2 to 3 times.

[0027]

According to the orthogonal transform coding method according to the present invention, an analysis window for windowing an input signal (so-called windowing) is provided at both ends with a width wider than an overlapping portion with an adjacent block. By using the one having the shoulder, it is possible to concentrate the spectrum on the frequency axis after the orthogonal transform, and realize highly accurate encoding with a small number of bits. In this case, since the width of the overlapping portion is not increased, the redundancy due to the overlapping between adjacent blocks does not increase, and the bit rate does not increase.

Further, according to the inverse orthogonal transform decoding apparatus of the present invention, after the analysis window having a shoulder wider at both ends than the overlap with the block adjacent on the time axis is applied, the orthogonal orthogonal transform is performed. What is claimed is: 1. An apparatus for performing inverse orthogonal transform decoding on a signal on a frequency axis which has been transformed and encoded, wherein a window obtained by multiplying the signal by the window function of the analysis window becomes 1 when added to an overlapping portion of an adjacent block. A function synthesis window is applied to the signal subjected to the inverse orthogonal transform, and the signal subjected to the synthesis window is connected to a signal of an adjacent block to restore the original signal on the time axis. Therefore, the coding efficiency can be improved without adversely affecting the waveform connection and the like.

[Brief description of the drawings]

FIG. 1 is a diagram showing an analysis window, a synthesis window, and a function shape obtained by multiplying the analysis window and the synthesis window for explaining an embodiment of the present invention.

FIG. 2 is a block circuit diagram illustrating a schematic configuration of an orthogonal transform decoding apparatus according to an embodiment of the present invention.

FIG. 3 is a block circuit diagram illustrating a schematic configuration of an inverse orthogonal transform decoding device according to an embodiment of the present invention.

FIG. 4 is a diagram illustrating a spectrum distribution on a frequency axis after orthogonal transformation according to an embodiment of the present invention.

FIG. 5 is a diagram showing a spectrum distribution on a frequency axis after orthogonal transformation in the related art.

FIG. 6 is a block circuit diagram showing a schematic configuration of a specific example of a band division adaptive transform coding apparatus to which an embodiment of the present invention can be applied.

FIG. 7 is a diagram illustrating a specific example of band division and blocking on the time axis for each band in the apparatus of FIG. 6;

FIG. 8 is a time chart for explaining windowing processing on an input signal.

FIG. 9 is a diagram showing each processing procedure of orthogonal transform coding and inverse orthogonal transform decoding.

FIG. 10 is a diagram illustrating a specific example of an analysis window and a synthesis window used in conventional orthogonal transform coding and inverse orthogonal transform decoding, and a function shape obtained by multiplying the analysis window and the synthesis window.

[Explanation of symbols]

 12 Blocking memory circuit 13 Analysis windowing circuit 14 Analysis window generation circuit 15 FFT Circuit 16: Encoder 22: Decoder 23: IFFT circuit 24: Composite windowing circuit 25 ········ Synthesis window generating circuit 26 ······ Waveform connection circuit

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-61-201526 (JP, A) JP-A-3-263925 (JP, A) International publication 90/14719 (WO, A1) IEICE technical research Report, Vol. 84, no. 330, CS84-181, "Sequencing and Evaluation of Window Functions", Ishii et al., March 1985, published by The Institute of Electronics, Information and Communication Engineers. (58) Fields surveyed (Int. Cl. ⁷ , DB name) H03M 7 / 30

Claims (2)

(57) [Claims] 1. An orthogonal transform coding method for applying an analysis window to an input signal to form a block on a time axis, and performing orthogonal transform for each block to encode the input signal. Is characterized by using a window function that has a shoulder with a width wider than the overlap with the adjacent block on the time axis, and concentrates the spectrum on the frequency axis after the orthogonal transform. Method. 2. A signal on a frequency axis which is orthogonally transformed and coded after an analysis window having a shoulder wider than an overlapping portion with a block adjacent on the time axis is applied to both ends, A window function such that the input signal is subjected to inverse orthogonal transform to be converted into a signal on the time axis, and the result of multiplication with the window function of the analysis window becomes 1 when the result of addition with the overlapping portion of the adjacent block is added. The composite window is applied to the signal subjected to the inverse orthogonal transformation, and the composite window is connected to a signal of an adjacent block to restore the original signal on the time axis. Inverse orthogonal transform decoding device.