CN1731694A - Digital audio frequency coding method and device - Google Patents
Digital audio frequency coding method and device Download PDFInfo
- Publication number
- CN1731694A CN1731694A CN 200410053479 CN200410053479A CN1731694A CN 1731694 A CN1731694 A CN 1731694A CN 200410053479 CN200410053479 CN 200410053479 CN 200410053479 A CN200410053479 A CN 200410053479A CN 1731694 A CN1731694 A CN 1731694A
- Authority
- CN
- China
- Prior art keywords
- mentioned
- subband
- digital audio
- bit
- bit value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention relates to an apparatus and digital audio coding method for high-speed calculation. The inventive method comprises steps of: calculating signal-to-masking (SMR) ratio with the proportional coefficient detected from digital audio signal and absolute masking value constant; according the division calculation value between the difference of SMR and target NMR (Noise-to-Masking Ratio) and constants to allocate bits value; quantization processing on digital audio signal according to the allocated bits value and generating frame. The invention has the advantages of compressing digital audio signal at high speed.
Description
Technical field
What relate in this invention is the digital audio coding method and the device that can carry out calculation at a high speed.
Background technology
Along with the extensive use of digital audio in various fields such as communication, computer, household electrical appliances, the problem of how to preserve and transmitting mass data has appearred.In order to solve such problem, determined the affiliated Moving Picture Experts Group-1 case of the International Organization for Stand in 91 years; By 94 years, on the basis of former Moving Picture Experts Group-1 case, expand and determined the Moving Picture Experts Group-2 case.
In this mpeg standard case, selected MUSICAM (the Masking-patternadapted Universal Subband Integrated Coding And Multiplexing) mode that in 128Kbps, can realize CD (Compact Disc) level tonequality as audio coding mode standard case.
To being described below of above-mentioned MUSICAM mode: utilizing auditory properties is a plurality of frequency bands territory (subband) with audio segmentation, and in each subband, according to its feature, utilization can be carried out the sub-band coding mode of quantification/encoding process audio frequency is carried out encoding process, with this but perception noise in each subband is controlled in the scope of minimizing, and in 96-128Kbps, realizes the recovery sound identical with former sound; It is chosen to be the 1st layer and the 2nd layer of MPEG.
On the one hand, the mpeg audio compress mode, the application purpose according to coding generally is divided into MPEG-1 and MPEG-2; According to bit rate, be divided into the 1st layer, the 2nd layer, the 3rd layer.
Here, the basic operation rule of MPEG-1 audio coding mode and MPEG-2 audio coding mode is identical, but the MPEG-1 audio coding mode can only carry out encoding process to two sound channels (left and right) at most, therefore acoustic image location (Sound Image Localization) is not very stable, and lacks presence.
In the MPEG-2 audio coding mode in order to embody this presence, on a left side (L), also increased center channel (C) on the basis of right (R) two channel stereo signal, surround channel (LC, RS) and low frequency special efficacy (Low Frequency Effect), and with SMPTE (Societyof Motion Picture and Television Engineers: film and TV tech NAB National Association 0f Broadcasters), the 5.1 sound channels coding algorithm that is made of 3/2+1 in ITU (International Telecommunication Union, international electric communication associating) the advice literary composition has been carried out standardization.
Fig. 1 is the 1st, the 2 layers of code device basic comprising figure of level that meet common mpeg standard.
With reference to Fig. 1, general code device comprises following parts: Methods of Subband Filter Banks 100, and carry out digitized 768Kbps/ch digital audio and video signals with the 48kHz/16 bit and be split into 32 subbands; Proportionality coefficient checkout gear 101 is exported and is divided into the digital audio and video signals of 32 subbands from above-mentioned Methods of Subband Filter Banks 100, detects the proportionality coefficient of each subband; FFT (FastFourier Transform) device 102 carries out Fourier transform to the 768kbps/ch digital audio and video signals and handles, and exports its frequency spectrum; Letter is covered than calculation element 103, the maximum spectrum of relatively selecting each subband between the proportionality coefficient that frequency spectrum and aforementioned proportion coefficient checkout gear 101 by 102 outputs of above-mentioned FFT device detects, utilize with the corresponding signal power source of above-mentioned maximum spectrum (signal power) then and cover threshold values (masking threshold) calculating letter and cover than (SMR:Signal-to-Masking Ratio is hereinafter to be referred as SMR); Bit Allocation in Discrete device 104, utilize the letter that calculates in the above-mentioned psychological acoustics model cover than and signal to noise ratio (SNR:Signal-to-Noise Ratio, be designated hereinafter simply as SNR) calculate noise masking than (NMR:Noise-to-Masking Ratio, be designated hereinafter simply as NMR), carry out Bit Allocation in Discrete according to the noise masking ratio that calculates then; Quantization device 105, the digital audio and video signals of above-mentioned Methods of Subband Filter Banks 100 being exported according to the bit value of above-mentioned Bit Allocation in Discrete device distribution carries out quantification treatment; Bit column generating apparatus 106 in the digital audio and video signals that has quantized, generates the frame that comprises additional information in above-mentioned quantization device 105.Here, above-mentioned FFT device and letter are covered than calculation element and are applicable to psychological acoustics model.And, comprise in the additional information here that last recovery quantizes/required proportionality coefficient index information and bit distribution information etc. during the compression digital audio frequency signal.
More particularly, in order to remove statistics repeatability, utilize the equidistant bank of filters 100 of 32 WOLA (WeightedOverlap Add (weighted superposition)) mode to convert the digital audio and video signals of importing to the subband sample, and in psychological acoustics (Psychoacoustic) model that uses fast Fourier transform, remove perceptual redundancy, try to achieve then and cover threshold values, and the bit distribution information that is applicable to the quantification treatment process is provided.
In addition, use the single Methods of Subband Filter Banks 100 of WOLA mode of 32 identical sizes in the 1st, 2 layer of MPEG, the filter that is used for Substrip analysis is a standard with the 512-tap low pass filter; Because the ranks calculation, occurrence frequency moves, and is split into the subband of 32 identical sizes.
The psychological acoustics model that is provided by MPEG is pure tone (Tonal) composition and noise (Non-tonal) composition with the FFT spectrum division, and considering under the prerequisite of definitely covering limit (AbsoluteMasking Threshold) factor, calculate the threshold values that covers of each composition.Among the 2nd layer of the MPEG, tell about, utilize the psychological acoustics Model Calculation to cover threshold values, and use during than little compressible at needs as the front.
Usually, the SMR that calculates in the psychological acoustics model is as the ratio between the signal power source of covering threshold values (Masking Threshold) and calculating from FFT frequency spectrum and proportionality coefficient of psychological acoustics model result; If during with decibel (dB) expression, just can be expressed as following mathematical expression 1.
[mathematical expression 1]
SMR (dB)=signal power source (dB)-cover threshold values (dB)
Try to achieve 32 SMR by each subband, each frame here.The physical significance of SMR is that by each subband, the signal power source value is than covering the big relatively degree of threshold values.
Here cover threshold values, tell about as the front, with the FFT spectrum division is pure tone (Tonal) composition and noise (Non-tonal) composition, and is considering that each composition that calculates covers threshold values under the prerequisite of definitely covering limit (Absolute Masking Threshold) factor.At this moment, definitely covering limit, is the demonstration curve with the digital audio and video signals frequency spectrum of input frequency band territory bottom line size irrelevant, that just can hear by normal hearing; Be finally to cover the value that is reflected in the process of threshold values trying to achieve psychological acoustics model.
Fig. 2 is the schematic diagram that shows 0 each subband SMR curve according to the particular frame among Fig. 1.
As shown in Figure 2, SMR not only between subband 1 and subband 17, have more than the 0dB on the occasion of, but also between subband 18 and subband 32, have negative value below the 0dB.At this moment, have the subband interval (for example, between subband 18 and the subband 32) of the following negative value of 0dB, all signals all are in shielding status, and this shows need not carry out Bit Allocation in Discrete again.Therefore only to have more than the 0dB on the occasion of subband interval (for example, between subband 1 and the subband 17) carry out Bit Allocation in Discrete.
On the one hand, in order to detect the proportionality coefficient that each subband sample value normalization (normalization) is handled, at first in the regular absolute value of 12 samples, find out maximum.Then 64 proportionality coefficients among this maximum and the MPEG are compared; According to comparative result, the proportionality coefficient of this frame will only be defined as less than the peaked proportionality coefficient of normalization.
Above-mentioned Bit Allocation in Discrete device in 32 divided subbands, is given allocation of subbands 1 bit of NMR calculated value maximum earlier; Then recomputate the NMR of each subband, and again to maximum allocation of subbands 1 bit of NMR value; Total bit number that the carrying out repeatedly of said process is assigned in number of times and the frame is identical.
The NMR that uses in above-mentioned bit allocation procedures utilizes SNR and SMR can be shown as following mathematical expression 2.
[mathematical expression 2]
NMR(dB)＝SMR(dB)-SNR(dB)
Here, SNR is the quantification noise that takes place in the quantizing process and the ratio between the source signal power values.
The physical significance of above-mentioned NMR is, it is big relatively that the quantification noise ratio of relevant subbands covers minimum value.In other words, the noise removed of main point is just many more more for NMR.
Therefore, during Bit Allocation in Discrete, the bit that the subband that NMR is big more need distribute is just many more.Usually per minute is joined 1 bit, and SNR will improve 6dB.It seems that so so-called Bit Allocation in Discrete is exactly by the suitable bit of allocation of subbands, the NMR value all to be converted into negative, so that quantize noise less than the processing procedure of covering threshold values.
If like this, normal person's the sense of hearing is to hear the noise that produces in the cataloged procedure, therefore can compress under the prerequisite that does not have the tonequality loss.
In the mathematical expression 2, SMR is according to the signal power source value and covers the fixed value that threshold values calculates that generally speaking SNR plays a key effect in this mathematical expression.
Fig. 3 is the schematic diagram that shows the NMR curvilinear motion according to the Bit Allocation in Discrete among Fig. 1.
In general, be in the state that distributes 0 bit on all subbands before the Bit Allocation in Discrete, so all signals all there is noise; At this moment SNR is 0dB.In other words, the NMR at initial stage identical with SMR (a).
(a) curve display is, all signals all have noise before the Bit Allocation in Discrete, and SNR=0dB, NMR=SMR.As (a) curve display, subband with 0dB be standard be divided into have on the occasion of subband interval (between subband 1 and the subband 17) and the subband interval (between subband 18 and the subband 32) with negative value.Under this situation, as telling about the front, the subband interval that has negative value below the 0dB be covered, need not be in the interval of Bit Allocation in Discrete, only to have more than the 0dB on the occasion of the subband interval carry out Bit Allocation in Discrete, NMR is adjusted into below the 0dB.
As shown in Figure 3, based on (a) curve,, all can be adjusted into 0dB following (b) in all subband intervals (c) by suitable Bit Allocation in Discrete.At this moment the bit allocation procedures of Shi Yonging is told about content as the front.
Bit rate just is (b) when being suitable for the bit number of present signal encoding, and and then 0dB forms the NMR value; But bit rate is abundant when high (c), and till remaining bits all was assigned with, NMR value continuation reduced, so the NMR value will be distributed in and is lower than the more position of 0dB.
When bit rate is higher as (c) curve, there are not much meanings here, as its accuracy of SMR of psychological acoustics model result.In other words, bit rate is abundant when high, even SMR goes up the error that exists to a certain degree, people's normal hearing is the quantification noise in the time of can't hearing Bit Allocation in Discrete.
But existing digital audio encoding apparatus is not considered this situation at all, just unconditionally tries to achieve by psychological acoustics model and covers threshold values, and calculate SMR according to the threshold values that covers of trying to achieve.And, be basic calculation NMR with above-mentioned SMR, by cyclic process repeatedly NMR is carried out Bit Allocation in Discrete again.Therefore, existing digital audio encoding apparatus needs the surmount function calculation of a lot of similar indexes or logarithm etc. in the SMR of psychological acoustics model computational process; And bit allocation procedures lasts till repeatedly always and finishes till the noise shaped (noise shaping), therefore still needs sizable calculation amount.Particularly, when bit rate is high, consume the more time of implementation.
In order to help to understand, illustrate below.If the calculation amount of decoding device is when embodiment is stereo, approximately need 10-20MIPS (Million Instructions Per Second (million instructions per second)), and existing code device is embodying extensive type DSP (DigitalSignal Processing: in the time of Digital Signal Processing), need the calculation amount of 80MIPS.And 70% from psychological acoustics model and the Bit Allocation in Discrete calculation in the calculation amount of above-mentioned existing code device.
Summary of the invention
Purpose of the present invention will solve above-mentioned variety of problems exactly; By digital audio coding method provided by the invention and device, when omitting psychological acoustics model, utilize brand-new mode to carry out Bit Allocation in Discrete, thereby can carry out the high speed calculation.
In order to achieve the above object, digital audio coding method provided by the invention comprises following several stages: in the stage one, digital audio and video signals is divided into most subbands; In the stage two, cut apart the proportionality coefficient that digital audio and video signals detects each subband from above-mentioned; In the stage three, utilize the above-mentioned proportionality coefficient that is detected and definitely cover the letter that threshold values calculates each subband and cover ratio; In the stage four, the letter that utilizes aforementioned calculation to come out is covered ratio, carries out Bit Allocation in Discrete by subband; In the stage five, according to the above-mentioned paricular value of distribution ratio, the digital audio and video signals that will cut apart by above-mentioned subband carries out quantification treatment; In the stage six, above-mentionedly quantized in the digital audio and video signals to generate the frame bit column that comprises additional information.
Above-mentioned digital audio coding method also comprises following two stages: 1 stage, total bit value of allowing by the bit of the allocation of subbands sum total and the frame of an above-mentioned majority subband formation deduct decide bit value difference compare; 2 stages, total bit value of allowing by the bit of allocation of subbands sum total and above-mentioned frame deduct decide the difference of bit value when inconsistent, the above-mentioned bit of having crossed by allocation of subbands is distributed again.
In addition, above-mentioned digital audio coding method also comprised with the next stage: the letter that calculates by subband cover than in selected letter in being present between institute's stator zone cover ratio, and carry out amplification processing to a certain degree.
Digital audio encoding apparatus according to the present invention is made up of following parts: the device that digital audio and video signals is divided into a plurality of subbands; Cut apart the device that digital audio and video signals detects each subband proportionality coefficient from above-mentioned; Utilize the above-mentioned proportionality coefficient that is detected and definitely cover threshold values calculate each subband letter cover than device; The letter that utilizes aforementioned calculation to come out is covered ratio, carries out the device of Bit Allocation in Discrete by subband; According to the above-mentioned paricular value of distribution ratio, the digital audio and video signals that will cut apart by above-mentioned subband carries out the device of quantification treatment; The above-mentioned device that has quantized to generate in the digital audio and video signals frame bit column that comprises additional information.
Tell about as the front, digital audio coding method in this invention and digital audio decoding device do not use existing psychological acoustics model, and direct proportion of utilization coefficient and definitely cover threshold values calculating SMR value; Utilize the SMR value of calculating then, and carry out Bit Allocation in Discrete according to the simple mathematical formula; Thereby avoided as in the prior art by psychological acoustics model and repeatedly the calculation amount that causes of computing increase phenomenon, and significantly reduced associated hardware costs, and can high speed compression digital audio frequency signal.
Illustrate, according to the coding method in this invention, about 10-20MIPS (stereo standard) (left and right) can encode among the extensive use type DSP, this calculation amount is equivalent to the calculation amount of common decoder level, compared with needing the existing encoder about 80MIPS to compare very large superiority.
Therefore, this invention is suitable for making that the execution time in this field significantly reduces, and having reduced hardware costs greatly in the application that PVR (Personal Video Recorder) etc. preserves with high bit rate.In addition, the portable type audio frequency apparatus that this invention can also be applied to require low Electric Design (for example, MP3 etc.) or need in multichannel (more than the 3 sound channels) audio frequency apparatus of a large amount of calculation amounts, only need about 30% calculation amount of prior art just can reach corresponding requirement.
Description of drawings
Fig. 1 is the 1st, the 2 layers of code device basic comprising figure of level that meet common mpeg standard.
Fig. 2 is for showing the schematic diagram of each subband SMR curve according to the particular frame among Fig. 1.
Fig. 3 is the schematic diagram that shows the NMR curvilinear motion according to the Bit Allocation in Discrete among Fig. 1.
Fig. 4 is for showing the skeleton diagram that digital audio encoding apparatus constitutes by this invention example.
Fig. 5 is for showing the schematic diagram of each subband SMR curve according to the particular frame in this invention example.
Fig. 6 is according to the SMR in this invention example, shows the schematic diagram of Bit Allocation in Discrete.
Accompanying drawing major part symbol description:
11: Methods of Subband Filter Banks 12: the proportionality coefficient checkout gear
13:SMR calculation element 14: Bit Allocation in Discrete device
15: quantization device 16: the bit column generating apparatus
Embodiment
With reference to the accompanying drawings, digital audio coding method in this invention and device are described in detail.
Fig. 4 shows the skeleton diagram that digital audio encoding apparatus constitutes by this invention example.
With reference to Fig. 4, the digital audio encoding apparatus in this invention comprises following parts: Methods of Subband Filter Banks 11 is divided into most subbands with digital audio and video signals; Proportionality coefficient checkout gear 12 from the digital audio and video signals that above-mentioned Methods of Subband Filter Banks 11 is cut apart, detects the proportionality coefficient of each subband; SMR calculation element 13 utilizes the proportionality coefficient that aforementioned proportion coefficient checkout gear 12 detects and definitely covers threshold values to calculate letter and cover ratio; Bit Allocation in Discrete device 14, the letter that utilizes above-mentioned SMR calculation element 13 to calculate is covered ratio, carries out Bit Allocation in Discrete by subband; Quantization device 15, the bit value according to above-mentioned Bit Allocation in Discrete device 14 distributes carries out quantification treatment to the above-mentioned digital audio and video signals of cutting apart by subband; Bit column generating apparatus 16 in the digital audio and video signals that has quantized, generates the frame that comprises additional information in above-mentioned quantization device 15.The index information and the bit distribution information that comprise proportionality coefficient here, in the said additional information.
Above-mentioned Methods of Subband Filter Banks 11 is utilized the uniformly-spaced bank of filters of 32 WOLA (Weighted Overlap-Add) mode in order to remove statistics repeatability, and the digital audio and video signals of importing is divided into 32 subbands.
Aforementioned proportion coefficient checkout gear 12 is at first found out maximum in order to detect the proportionality coefficient that each subband sample value normalization (normalization) is handled in the regular absolute value of 12 samples.Then 64 proportionality coefficients among this maximum and the MPEG are compared; According to comparative result, the proportionality coefficient of this frame will only be defined as less than the peaked proportionality coefficient of normalization.At this moment the proportionality coefficient that detects just is defined as the signal power source value of respective sub-bands.
In the application of prior art, choose the big signal power source value that is defined as respective sub-bands in proportionality coefficient and the FFT frequency spectrum power supply; But, unconditionally proportionality coefficient is defined as the signal power source value of respective sub-bands in this invention.This be because, tell about as the front, FFT frequency spectrum power values and proportionality coefficient as psychological acoustics model result compare, get big value defined signal power source value according to comparative result then, and the SMR that the signal power source value of SMR that calculates by this signal power source value and the definition of passing ratio coefficient is calculated there is not any difference.
Therefore, dispensed the psychological acoustics model that needs complicated calculation amounts in this invention, and proportionality coefficient that aforementioned proportion coefficient checkout gear is detected directly is utilized as the signal power source value, calculates SMR with this.
The proportionality coefficient that above-mentioned SMR calculation element 13 detects the proportionality coefficient checkout gear is utilized as the signal power source value and calculates SMR.At this moment, SMR can be expressed as following mathematical expression 3.
[mathematical expression 3]
SMR (dB)=proportionality coefficient (dB)-definitely covers threshold values (dB)
Here, as the front tell about, proportionality coefficient is the signal power source value of trying to achieve from aforementioned proportion coefficient checkout gear.In addition, definitely cover threshold values and be and realize the fixed constant value calculated by experiment, and each subband all there is its fixed constant.Above-mentioned what cover definitely that threshold values represents is the frequency field bottom line size that normal hearing irrelevant with the digital audio and video signals frequency spectrum, the people can be heard, is the value of calculating by experiment.
Therefore, do not use the psychological acoustics model that needs complicated calculation amount in this invention, just utilize and calculate the simple proportional coefficient and calculate SMR as the threshold values that definitely covers of constant.
By above-mentioned SMR calculation element 13, the SMR curve that calculates by subband as shown in Figure 5.
Fig. 5 is for showing the schematic diagram of each subband SMR curve according to the particular frame in this invention example.Here, be the SMR curve that utilizes existing psychological acoustics Model Calculation to come out (a), be to utilize the proportionality coefficient in this invention and definitely cover the SMR curve that threshold values calculates (b).
As shown in Figure 5, the SMR curve (b) in this invention is lower to a certain degree than existing SMR curve (a).Though how many SMR curves (b) in this invention exists some errors than the existing SMR curve (a) that approaches actual value, see still to be more similar on the whole.But above-mentioned error, can be described as inappreciable at the time of wasting in the utilization compared with the psychological acoustics model of complexity.
On the one hand, the SMR that Bit Allocation in Discrete device 14 utilizes the SMR calculation element to calculate carries out Bit Allocation in Discrete.
In the application of prior art, the total bit number that is assigned in Bit Allocation in Discrete number of times and each frame is suitable, the problem that the calculation amount increases therefore occurs; In order to address the above problem, adopted new Bit distribution method in this invention.
At first carry out Bit Allocation in Discrete, can be expressed as following mathematical expression 4 by subband.
[mathematical expression 4]
Bitalloc (sb)=[(SMR (sb)-target NMR)/6]
But (SMR (sb)-target NMR)/6 show as at 6 o'clock, and [a] is the minimum positive number that is not less than a.
Here, with 6 reasons of removing difference between SMR value and the target NMR value be exactly, per minute is joined 1 bit, the SNR 6dB that will make progress.Therefore, when utilizing above-mentioned mathematical expression 6 to carry out Bit Allocation in Discrete, the upwards value of SNR is identical with allocation bit last 6 value on duty, and the NMR value of calculating thus can be reduced to negative value.
At this moment, target NMR is according to the fixing constant value of coding bit rate, and the high more NMR value of bit rate can be more little; Preferably little than 0dB.
For example, the SMR of subband 1 be the SMR of 20dB, subband 2 be 27dB, target NMR be-during 30dB,, distribute [(20-(30))/6]=[50/6]=[8.33]=9 bit in the subband 1 according to mathematical expression.In addition, distribute [(27-(30))/6]=[57/6]=[9.33]=10 bit in the subband 2.Utilize said process, carry out Bit Allocation in Discrete by subband.
But, to tell about as the front, not all subband all wants allocation bit.In other words, the subband with the following SMR of 0dB just need not carry out Bit Allocation in Discrete.
It should be noted that according to mathematical expression 4 distribute with each subband in the bit value sum total may be bigger or little than total bit value that a frame that comprises all subbands is allowed.At this moment, the smaller sum total of total bit value of preferably allowing than a frame distribute with each subband in.In addition, being allocated in the bit value sum total of each subband also can be consistent with total bit value that a frame is allowed.
Above-mentioned these contents can be expressed as following mathematical expression 5.
[mathematical expression 5]
Total bit value that total-bits=allows by bit value sum total≤each frame of allocation of subbands-decide bit value
Above-mentioned Bit Allocation in Discrete device 14 according to above-mentioned mathematical expression 4 to each allocation of subbands subband, and the bit sum total that will be assigned to each subband and every frame total bit value of allowing deduct decide bit value difference compare.
Comparative result, total bit value of allowing by the bit of allocation of subbands sum total and every frame deduct decide the difference of bit value when inconsistent, redistribute the bit value that is assigned in each subband.
In other words, when summing up total bit value of allowing, deduct the bit value that exceeds the bit value from being assigned with of each subband greater than every frame by the bit value of allocation of subbands.
On the contrary, when summing up total bit value of allowing, add not enough bit value in the bit value to being assigned with of each subband less than every frame by the bit value of allocation of subbands.
As, redistribute by subband in the process of bit value, if the bit value subtraction, just from the highest subband; If the bit value additional calculation, just from minimum subband.
Fig. 6 is according to the SMR in this invention example, shows the schematic diagram of Bit Allocation in Discrete.
(a) is that SMR curve, (b) in this invention are that SMR curve, (c) during prior art is used is to be that bit value curve, (d) that standard is distributed are that final NMR curve, (e) that Bit Allocation in Discrete with (c) obtains when being adapted to (a) SMR curve are the final NMR curves that Bit Allocation in Discrete with (c) shows when being adapted to (b) SMR with (a) among Fig. 6.
As shown in Figure 6, the SMR curve (a) in this invention is carried out (c) after the Bit Allocation in Discrete by subband, try to achieve NMR curve (b) with the above-mentioned bit value that is assigned with; And the bit value (c) in suitable this invention in the SMR curve (b) in the prior art application, try to achieve NMR curve (e).As shown in Figure 6, when being adapted to the Bit Allocation in Discrete in this invention in the prior art SMR curve (e), (e) also show all NMR curves below 0dB of all values, so the NMR curve (d) among the present invention is very effective.In addition, though (d) curve need only the bit rate height, and difference is more abundant, just can not have any problem between the two with (e) curve is different mutually.
We can learn in Fig. 6, even the psychological acoustics model of unfavorable usefulness also can fully hide the quantification noise, can guarantee not have the cataloged procedure of tonequality damage.
When each allocation of subbands behind bit value, above-mentioned quantization device 15 is according to the bit value that is assigned with in each subband, and the digital audio and video signals that is assigned with of each subband is carried out quantification treatment.
The above-mentioned digital audio and video signals that quantized is generated as the frame bit column that includes additional information under the effect of above-mentioned bit column generating apparatus 16.
In fact, the abundant experimental result of multiple source of sound is shown, the SMR value among the present invention and have now between the SMR value difference as shown in Figure 5, at subband 10 to the error that takes place between the subband 20 about maximum 20dB.
But, as long as to be present in error takes place subband 10 between the subband 20 and by the SMR value that this invention is tried to achieve, carry out corresponding to the processing of the amplification of error amount, above-mentioned error problem will simply solve.
Therefore, carry out before the Bit Allocation in Discrete in the above-mentioned Bit Allocation in Discrete device 14, can amplification subband 10 to the SMR value between the subband 20.
The amplification of the interval SMR value of this particular sub-band is not necessary to carry out, and just carries out getting final product when being necessary.
Claims (15)
1, a kind of being used for compressed the digital audio coding method of input digital audio signal at a high speed, it is characterized in that comprising the following several stages content:
In the stage one, digital audio and video signals is divided into most subbands;
In the stage two, cut apart the proportionality coefficient that digital audio and video signals detects each subband from above-mentioned;
In the stage three, utilize the above-mentioned proportionality coefficient that is detected and definitely cover the letter that threshold values calculates each subband and cover ratio;
In the stage four, the letter that utilizes aforementioned calculation to come out is covered ratio, carries out Bit Allocation in Discrete by subband;
In the stage five, according to the above-mentioned paricular value of distribution ratio, the digital audio and video signals that will cut apart by above-mentioned subband carries out quantification treatment;
In the stage six, above-mentionedly quantized in the digital audio and video signals to generate the frame bit column that comprises additional information.
2, according to the described digital audio coding method of claim 1, it is characterized in that: the described bit that is assigned with calculates according to following mathematical expression,
Bitalloc (sb)=[(SMR (sb)-target NMR)/6]
But (SMR (sb)-target NMR)/6 show as at 6 o'clock, and [a] is the minimum positive number that is not less than a;
Bitalloc is the bit value that is assigned with; SMR is that letter is covered ratio; Target NMR is the constant value of fixing according to coding bit rate.
3, according to the described digital audio coding method of claim 1, it is characterized in that also comprising:
1 stage, total bit value of allowing by the bit of the allocation of subbands sum total and the frame of an above-mentioned majority subband formation deduct decide bit value difference compare;
2 stages, total bit value of allowing by the bit of allocation of subbands sum total and above-mentioned frame deduct decide the difference of bit value when inconsistent, the above-mentioned bit of having crossed by allocation of subbands is distributed again.
4, according to the described digital audio coding method of claim 3, it is characterized in that: when above-mentioned bit value by allocation of subbands is summed up total bit value of allowing above above-mentioned every frame, deduct the above-mentioned bit value part that exceeds in the bit value that from above-mentioned subband, is assigned to.
5, according to the described digital audio coding method of claim 3, it is characterized in that: when above-mentioned bit value by allocation of subbands is summed up total bit value of allowing less than above-mentioned every frame, add above-mentioned not enough bit value part in the bit value that in above-mentioned height band, is assigned to.
6, according to the described digital audio coding method of claim 3, it is characterized in that: redistribute by subband in the process of bit value, if the bit value subtraction, just from the highest subband; If the bit value additional calculation, just from minimum subband.
7,, it is characterized in that also comprising with the next stage according to the described digital audio coding method of claim 1: the letter that calculates by subband cover than in selected letter in being present between institute's stator zone cover ratio, and carry out amplification processing to a certain degree.
8, according to the described digital audio coding method of claim 7, it is characterized in that: above-mentioned subband interval is that subband (10) is between the subband (20).
9, a kind of digital audio encoder is characterized in that comprising:
Digital audio and video signals is divided into the device of most subbands;
Cut apart the device that digital audio and video signals detects each subband proportionality coefficient from above-mentioned;
Utilize the above-mentioned proportionality coefficient that is detected and definitely cover threshold values calculate each subband letter cover than device;
The letter that utilizes aforementioned calculation to come out is covered ratio, carries out the device of Bit Allocation in Discrete by subband;
According to the above-mentioned paricular value of distribution ratio, the digital audio and video signals that will cut apart by above-mentioned subband carries out the device of quantification treatment; And
The above-mentioned device that has quantized to generate in the digital audio and video signals frame bit column that comprises additional information.
10, according to the described digital audio encoder of claim 9, it is characterized in that: above-mentioned Bit Allocation in Discrete device according to letter cover than and target noise hide than between difference remove the result of calculation value of permanent numerical value carry out Bit Allocation in Discrete.
11,, it is characterized in that also comprising following two devices according to the described digital audio encoder of claim 9:
Device 1, total bit value of allowing by the bit of the allocation of subbands sum total and the frame of an above-mentioned majority subband formation deduct decide bit value difference compare;
Device 2, total bit value of allowing by the bit of allocation of subbands sum total and above-mentioned frame deduct decide the difference of bit value when inconsistent, the above-mentioned bit of having crossed by allocation of subbands is distributed again.
12, according to the described digital audio encoding apparatus of claim 11, it is characterized in that: when above-mentioned bit value by allocation of subbands is summed up total bit value of allowing above above-mentioned every frame, deduct the above-mentioned bit value part that exceeds in the bit value that from above-mentioned subband, is assigned to.
13, according to the described digital audio encoding apparatus of claim 11, it is characterized in that: when above-mentioned bit value by allocation of subbands is summed up total bit value of allowing less than above-mentioned every frame, add above-mentioned not enough bit value part in the bit value that in above-mentioned height band, is assigned to.
14, according to the described digital audio encoder of claim 11, it is characterized in that: redistribute by subband in the process of bit value, if the bit value subtraction, just from the highest subband; If the bit value additional calculation, just from minimum subband.
15, according to the described digital audio encoder of claim 9, it is characterized in that also comprising: the letter that calculates by subband cover than in selected subband (10) to the interval letter of subband (20) that is present in cover ratio, and carry out the device that amplification is to a certain degree handled.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200410053479 CN1731694A (en) | 2004-08-04 | 2004-08-04 | Digital audio frequency coding method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200410053479 CN1731694A (en) | 2004-08-04 | 2004-08-04 | Digital audio frequency coding method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1731694A true CN1731694A (en) | 2006-02-08 |
Family
ID=35964005
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200410053479 Pending CN1731694A (en) | 2004-08-04 | 2004-08-04 | Digital audio frequency coding method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1731694A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101432804B (en) * | 2006-03-13 | 2013-01-16 | 法国电信公司 | Method of coding a source audio signal, corresponding coding device, decoding method and device |
| CN108449704A (en) * | 2013-10-22 | 2018-08-24 | 韩国电子通信研究院 | Generate the method and its parametrization device of the filter for audio signal |
-
2004
- 2004-08-04 CN CN 200410053479 patent/CN1731694A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101432804B (en) * | 2006-03-13 | 2013-01-16 | 法国电信公司 | Method of coding a source audio signal, corresponding coding device, decoding method and device |
| CN108449704A (en) * | 2013-10-22 | 2018-08-24 | 韩国电子通信研究院 | Generate the method and its parametrization device of the filter for audio signal |
| CN108449704B (en) * | 2013-10-22 | 2021-01-01 | 韩国电子通信研究院 | Method for generating a filter for an audio signal and parameterization device therefor |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1146203C (en) | Dynamic bit allocation apparatus and method for audio coding | |
| CN1153191C (en) | Scalable coding method for high quality audio | |
| EP2282310B1 (en) | Entropy coding by adapting coding between level and run-length/level modes | |
| CN1030129C (en) | High efficiency digital data encoding and decoding apparatus | |
| CN1154085C (en) | Scalable audio coding/decoding method and apparatus | |
| JP3762579B2 (en) | Digital audio signal encoding apparatus, digital audio signal encoding method, and medium on which digital audio signal encoding program is recorded | |
| US7136418B2 (en) | Scalable and perceptually ranked signal coding and decoding | |
| CN1092420C (en) | Audio data decoding apparatus and audio data coding/decoding system | |
| US9741351B2 (en) | Adaptive quantization noise filtering of decoded audio data | |
| CN1183685C (en) | System and method for entropy ercoding quantized transform coefficients of a sigral | |
| JP5162588B2 (en) | Speech coding system | |
| JP5400143B2 (en) | Factoring the overlapping transform into two block transforms | |
| CN1669072A (en) | Low bit-rate audio coding | |
| CN1119902A (en) | Method and apparatus for compressing, method for transmitting, and method and apparatus for expanding compressed multi-channel sound signals | |
| CN1675683A (en) | Device and method for scalable coding and device and method for scalable decoding | |
| CN101055720A (en) | Method and apparatus for encoding and decoding an audio signal | |
| CN1662958A (en) | Audio coding system using spectral hole filling | |
| CN1677490A (en) | Intensified audio-frequency coding-decoding device and method | |
| CN1262990C (en) | Audio coding method and apparatus using harmonic extraction | |
| CN1677491A (en) | Intensified audio-frequency coding-decoding device and method | |
| JP2001343997A (en) | Method and device for encoding digital acoustic signal and recording medium | |
| JP5970602B2 (en) | Audio encoding and decoding with conditional quantizer | |
| JP2003324355A (en) | Method and device for encoding, method and device for decoding, and program and recording medium | |
| CN1677492A (en) | Intensified audio-frequency coding-decoding device and method | |
| CN1485849A (en) | Digital audio encoder and its decoding method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |