WO2007008001A2 - Apparatus and method of encoding and decoding audio signal - Google Patents

Abstract

In one embodiment, a channel in a frame of the audio signal is subdivided into a plurality of blocks, and at least two of the blocks have different lengths. An optimum prediction order for each block is determined based on a permitted prediction order and a length of the block.

Description

[DESCRIPTION]

APPARATUS AND METHOD OF ENCODING AND DECODING AUDIO SIGNAL

BACKGROUND OF THE INVENTION

The present invention relates to a method for processing audio signal, and more particularly to a method and apparatus of

encoding and decoding audio signal.

The storage and replaying of audio signals has been accomplished in different ways in the past. For example,

music and speech have been recorded and preserved by

phonographic technology (e.g., record players), magnetic

technology (e.g., cassette tapes), and digital technology

(e.g., compact discs). As audio storage technology progresses,

many challenges need to be overcome to optimize the quality

and storability of audio signals.

For the archiving and broadband transmission of music signals,

lossless reconstruction is becoming a more important feature

than high efficiency in compression by means of perceptual

coding as defined in MPEG standards such as MP3 or AAC.

Although DVD audio and Super CD Audio include proprietary

lossless compression schemes, there is a demand for an open

and general compression scheme among content-holders and broadcasters. In response to this demand, a new lossless

coding scheme has been considered as an extension to the MPEG-

4 Audio standard. Lossless audio coding permits the compression of digital audio data without any loss in quality- due to a perfect reconstruction of the original signal.

SUMMARY OF THE INVENTION The present invention relates a method of processing an audio signal.

In one embodiment, a channel in a frame of the audio signal is subdivided into a plurality of blocks, and at least two of the blocks have different lengths. An optimum prediction order for each block is determined based on a permitted prediction order and a length of the block.

In one embodiment, the determination of the optimum prediction order includes determining a global prediction order based on the permitted prediction order, determining a local prediction order based on the length of the block, and selecting a minimum one of the global prediction order and the local prediction order as the optimum prediction order. For example, the global prediction order may be determined as equal to,

ceil (Iog2 (permitted prediction order +1) .

As another example, the local prediction order may be determined as equal to,

max (ceil (Iog2 ( (Nb>>3) -1) ), 1) where Nb is the length of the block.

According to one embodiment, the channel is subdivided according to a subdivision hierarchy. The subdivision

hierarchy has more than one level, and each level is associated with a different block length. For example, a

superordinate level of the subdivision hierarchy is associated

with a block length double a block length associated with a

subordinate level . In one embodiment, if a channel has a length of N, then the

channel is subdivided into the plurality of blocks such that

each block has a length of one of N/2, N/4, N/8, N/16 and N/32.

In one embodiment, information is generated such that a length

of the information depends on a number of levels in the subdivision hierarchy. For example, the information may be

generated such that the information includes a number of

information bits, and the information bits indicate the

subdivision of the channel into the blocks. More specifically,

each information bit may be associated with a level in the

subdivision hierarchy and may be associated with a block at

the associated level, and each information bit indicates

whether the associated block was subdivided.

In one embodiment, the method further includes predicting

current data samples in the channel based on previous data

samples . The number of the previous data samples used in the predicting is referred to as the prediction order. A residual

of the current data samples is obtained based on the predicted

data samples .

In one embodiment, the prediction is performed by

progressively increasing the prediction order to a desired prediction order as previous data samples become available .

For example, this progressive prediction process may be

performed for random access frames, which are frames to be

encoded such that previous frames are not necessary to decode

the frame.

The present invention further relates to methods and apparatuses for encoding an audio signal, and to methods and

apparatuses for decoding an audio signal.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are included to provide a

further understanding of the invention and are incorporated in

and constitute a part of this application, illustrate

embodiment (s) of the invention and together with the

description serve to explain the principle of the invention. In the drawings :

FIG. 1 is an example illustration of an encoder according to

an embodiment of the present invention.

FIG. 2 is an example illustration of a decoder according to an embodiment of the present invention. FIG. 3 is an example illustration of a bitstream structure of

a compressed M-channel file according to an embodiment of the

present invention.

FIG. 4 is an example illustration of a conceptual view of a

hierarchical block switching method according to an embodiment

of the present invention.

FIG. 5 is an example illustration of a block switching

examples and corresponding block switching information codes .

FIG. 6 is an example illustration of block switching methods

for a plurality of channel according to embodiments of the present invention.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

Reference will now be made in detail to the preferred

embodiments of the present invention, examples of which are

illustrated in the accompanying drawings. Wherever possible,

the same reference numbers will be used throughout the

drawings to refer to the same or like parts .

Prior to describing the present invention, it should be noted

that most terms disclosed in the present invention correspond

to general terms well known in the art, but some terms have

been selected by the applicant as necessary and will hereinafter be disclosed in the following description of the

present invention. Therefore, it is preferable that the terms defined by the applicant be understood on the basis of their meanings in the present invention.

In a lossless audio coding method, since the encoding process

has to be perfectly reversible without loss of information,

several parts of both encoder and decoder have to be implemented in a deterministic way.

Codec Structure

FIG. 1 is an example illustration of an encoder 1 according to the present invention.

A partitioning part 100 partitions the input audio data into

frames. Within one frame, each channel may be further

subdivided into blocks of audio samples for further processing.

A buffer 110 stores block and/or frame samples partitioned by the partitioning part 100.

A coefficient estimating part 120 estimates an optimum set of

coefficient values for each block. The number of coefficients,

i.e., the order of the predictor, can be adaptively chosen as

well. The coefficient estimating part 120 calculates a set of

parcor values for the block of digital audio data. The parcor

value indicates parcor representation of the predictor

coefficient. A quantizing part 130 quantizes the set of

parcor values .

A first entropy coding part 140 calculates parcor residual

values by subtracting an offset value from the parcor value, and encodes the parcor residual values using entropy codes

defined by entropy parameters, wherein the offset value and

the entropy parameters are chosen from an optimal table. The

optimal table is selected from a plurality of tables based on

a sampling rate of the block of digital audio data. The plurality of tables are predefined for a plurality of sampling

rate ranges, respectively, for optimal compression of the

digital audio data for transmission.

A coefficient converting part 150 converts the quantized

parcor values into linear predictive coding (LPC) coefficients.

A predictor 160 estimates current prediction values from the

previous original samples stored in the buffer 110 using the

linear predictive coding coefficients. A subtracter 170

calculates a prediction residual of the block of digital audio data using an original value of digital audio data stored in

the buffer 110 and a prediction value estimated in the predictor 160.

A second entropy coding part 180 codes the prediction residual

using different entropy codes and generates code indices . The

indices of the chosen codes will be transmitted as auxiliary

information. The second entropy coding part 180 may code the

prediction residual using one of two alternative coding

techniques having different complexities. One coding

technique is the well-known Golomb-Rice coding (herein after

simply "Rice code") method and the other is the well-known Block Gilbert-Moore Codes (herein after simply "BGMC") method.

Rice codes have low complexity yet are efficient. The BGMC arithmetic coding scheme offers even better compression at the

expense of a slightly increased complexity compared to Rice

codes .

Finally, a multiplexing part 190 multiplexes coded prediction

residual, code indices, coded parcor residual values, and

other additional information to form a compressed bitstream.

The encoder 1 also provides a cyclic redundancy check (CRC) checksum, which is supplied mainly for the decoder to verify

the decoded data. On the encoder side, the CRC can be used to

ensure that the compressed data are losslessly decodable.

Additional encoding options include flexible block switching

scheme, random access and joint channel coding. The encoder 1

may use these options to offer several compression levels with

different complexities. The joint channel coding is used to

exploit dependencies between channels of stereo or multi¬

channel signals. This can be achieved by coding the

difference between two channels in the segments where this

difference can be coded more efficiently than one of the

original channels. These encoding options will be described

in more detail below after a description of an example decoder

according to the present invention.

FIG. 2 is an example illustration of a decoder 2 according to

the present invention. More specially, FIG. 2 shows the lossless audio signal decoder which is significantly less

complex than the encoder, since no adaptation has to be

carried out.

A demultiplexing part 200 receives an audio signal and

demultiplexes a coded prediction residual of a block of digital audio data, code indices, coded parcor residual values

and other additional information. A first entropy decoding

part 210 decodes the parcor residual values using entropy

codes defined by entropy parameters and calculates a set of

parcor values by adding offset values to the decoded parcor residual values; wherein the offset value and the entropy

parameters are chosen from a table selected by the decoder

from a plurality of tables based on a sampling rate of the

block of digital audio data. A second entropy decoding part

220 decodes the demultiplexed coded prediction residual using

the code indices. A coefficient converting part 230 converts

the entropy decoded parcor value into LPC coefficients. A

predictor 240 estimates a prediction residual of the block of

digital audio data using the LPC coefficients. An adder 250

adds the decoded prediction residual to the estimated

prediction residual to obtain the original block of digital

audio data. An assembling part 260 assembles the decoded

block data into frame data. Therefore, the decoder 2 decodes the coded prediction residual

and the parcor residual values, converts the parcor residual values into LPC coefficients, and applies the inverse

prediction filter to calculate the lossless reconstruction

signal. The computational effort of the decoder 2 depends on

the prediction orders chosen by the encoder 1. In most cases,

real-time decoding is possible even on low-end systems.

FIG. 3 is an example illustration of a bitstream structure of

a compressed audio signal including a plurality of channels

(e.g., M channels) according to the present invention.

The bitstream consists of at least one audio frame including a

plurality of channels (e.g., M channels) . The "channels"

field in the bitstream configuration syntax (see Table 6

below) indicates the number of channels. Each channel is sub¬

divided into a plurality of blocks using the block switching

scheme according to present invention, which will be described in detail later. Each sub-divided block has a different size

and includes coding data according to the encoding of FIG.l.

For example, the coding data within a subdivided block

contains the code indices, the prediction order K, the

predictor coefficients, and the coded residual values. If

joint coding between channel pairs is used, the block

partition is identical for both channels, and blocks are

stored in an interleaved fashion. A "js_stereo" field in the

bitstream configuration syntax (Table 6) indicates whether

joint stereo (channel difference) is on or off, and a

"js_switch" field in the frame_data syntax (See Table 7 below) indicates whether joint stereo (channel difference) is

selected. Otherwise, the block partition for each channel is independent .

Hereinafter, the block switching, random access, prediction,

and entropy coding options previously mentioned will now be described in detail with reference to the accompanying

drawings and syntaxes that follow.

Block Switching

An aspect of the present invention relates to subdividing each channel into a plurality of blocks prior to using the actual

coding scheme. Hereinafter, the block partitioning (or

subdividing) method according to the present invention will be

referred to as a "block switching method" .

Hierarchical Block Switching

FIG. 4 is an example illustration of a conceptual view of a

hierarchical block switching method according to the present

invention. For example, FIG. 4 illustrates a method of

hierarchically subdividing one channel into 32 blocks. When a

plurality of channels is provided in a single frame, each

channel may be subdivided (or partitioned) to up to 32 blocks,

and the subdivided blocks for each channel configure a frame. Accordingly, the block switching method according to the

present invention is performed by the partitioning part 100 shown in FIG. 1. Furthermore, as described above, the

prediction and entropy coding are performed on the subdivided

block units.

In general, conventional Audio Lossless Coding (ALS)

includes a relatively simple block switching mechanism. Each channel of N samples is either encoded using one full length

block (N_B = N) or four blocks of length N_B = N/4 (e.g., 1:4

switching) , where the same block partition applies to all channels. Under some circumstances, this scheme may have some

limitations. For example, while only 1:1 or 1:4 switching may

be possible, different switching (e.g., 1:2, 1:8, and

combinations thereof) may be more efficient in some cases.

Also in conventional ALS, switching is performed identically

for all channels, although different channels may benefit from

different switching (which is especially true if the channels

are not correlated) .

Therefore, the block switching method according to embodiments

of the present invention provide relatively flexible block switching schemes, where each channel of a frame may be

hierarchically subdivided into a plurality of blocks. For

example, FIG. 4 illustrates a channel which can be

hierarchically subdivided to up to 32 blocks. Arbitrary

combinations of blocks with N_B = N, N/2, N/4, N/8, N/16, and

N/32 may be possible within a channel according to the

presented embodiments, as long as each block results from a subdivision of a superordinate block of double length. For

example, as illustrated in the example shown in FIG. 4, a

partition into N/4 + N/4 + N/2 may be possible, while a

partition into N/4 + N/2 + N/4 may not be possible (e.g., block switching examples shown in FIGs. 5 (e) and 5 described

below) . Stated another way, the channel is divided into the

plurality of blocks such that each block has a length equal to one of ,

N/ (m¹) for i = 1, 2, ... p,

where N is the length of the channel, m is an integer greater

than or equal to 2 , and p represents a number of the levels in the subdivision hierarchy.

Accordingly, in embodiments of the present invention, a

bitstream includes information indicating block switching

levels and information indicating block switching results .

Herein, the information related to block switching is included

in the syntax, which is used in the decoding process, described in detail below.

For example, settings are made so that a minimum block size

generated after the block switching process is N_B =N/32.

However, this setting is only an example for simplifying the

description of the present invention. Therefore, settings

according to the present invention are not limited to this setting.

More specifically, when the minimum block size is N_B =N/32,

this indicates that the block switching process has been

hierarchically performed 5 times, which is referred to as a

level 5 block switching. Alternatively, when the minimum

block size is N_B =N/16, this indicates that the block

switching process has been hierarchically performed 4 times,

which is referred to as a level 4 block switching. Similarly,

when the minimum block size is N_B =N/8, the block switching

process has been hierarchically performed 3 times, which is referred to as a level 3 block switching. And, when the

minimum block size is N_B =N/4, the block switching process has

been hierarchically performed 2 times, which is referred to as

a level 2 block switching. When the minimum block size is N_B

=N/2, the block switching process has been hierarchically

performed 1 time, which is referred to as a level 1 block

switching. Finally, when the minimum block size is N_B =N, the

hierarchical block switching process has not been performed,

which is referred to as a level 0 block switching.

In embodiments of the present invention, the information

indicating the block switching level will be referred to as a

first block switching information. For example, the first

block switching information may be represented by a 2-bit

"block_switching" field within the syntax shown in Table 6, which will be described in a later process. More specifically,

"block_switching = 00" signifies level 0, "block_switching =

01" signifies any one of level 1 to level 3, "block_switching

= 10" signifies level 4, and "block_switching = 11" signifies

level 5.

Additionally, information indicating the results of the block

switching performed for each hierarchical level in accordance

with the above-described block switching levels is referred to

in the embodiments as second block switching information.

Herein, the second block switching information may be

represented by a "bs_info" field which is expressed by any one of 8 bits, 16 bits, and 32 bits within the syntax shown in

Table 7. More specifically, if ⁿblock_switching = 01"

(signifying any one of level 1 to level 3) , ^λNbs_info" is

expressed as 8 bits. If "block_switching = 10" (signifying

level 4), "bs_info" is expressed as 16 bits. In other words,

up to 4 levels of block switching results may be indicated by

using 16 bits. Furthermore, if ^wblock_switching = 11"

(signifying level 5, "bs__info" is expressed as 32 bits. In

other words, up to 5 levels of block switching results may be

indicated by using 32 bits. Finally, if "block_switching =

00" (signifying that the block switching has not been

performed) , "bs_info" is not transmitted. This signifies that one channel configures one block.

The total number of bits being allocated for the second block switching information is decided based upon the level value of

the first block switching information. This may result in

reducing the final bit rate. The relation between the first

block switching information and the second block switching

information is briefly described in Table 1 below.

Table 1: Block switching levels.

Hereinafter, an embodiment of a method of configuring (or

mapping) each bit within the second block switching

information (bs__info) will now be described in detail.

The bs_info field may include up to 4 bytes in accordance with

the above-described embodiments. The mapping of bits with

respect to levels 1 to 5 may be [(0)1223333 44444444 55555555

55555555] . The first bit may be reserved for indicating independent or synchronous block switching, which is described in more detail below in the Independent/Synchronous Block

Switching section. FIGs. 5(a)-5(f) illustrate different block

switching examples for a channel where level 3 block switching

may take place. Therefore, in these examples, the minimum

block length is N_B = N/8, and the bs_info consists of one byte.

Starting from the maximum block length N_B = N, the bits of

bs_info are set if a block is further subdivided. For example,

in FIG. 5 (a) , there is no subdivision at all, thus "bs_info" is (0)000 0000. In FIG. 5 (b) , the frame is subdivided ((0)1...)

and the second block of length N/2 is further split ((0)101...)

into two blocks of length N/4; thus "bs_info" is (0)1010 0000. In FIG. 5(c), the frame is subdivided ((0)1...), and only the

first block of length N/2 is further split ((0)110...) into two

blocks of length N/4; thus "bs_info" is (0)1100 0000. In FIG. 5 (d) , the frame is subdivided ((0)1...), the first and second

blocks of length N/2 is further split ((0)111...) into two

blocks of length N/4, and only the second block of length N/4 is further split ((0)11101...) into two blocks of length N/8;

thus "bs_info" is (0)111 0100.

As discussed above, the examples in FIGs. 5 (e) and 5(f)

represent cases of block switching that are not permitted

because the N/2 block in FIG. 5 (e) and the first N/4 block in

FIG. 5(f) could not have been obtained by subdividing a block

of the previous level . Independent / Synchronous Block Switching

FIGs. 6 (a) - 6 (c) are example illustrations of block switching

according to embodiments of the present invention.

More specifically, FIG. 6 (a) illustrates an example where block switching has not been performed for channels 1, 2, and

3. FIG. 6 (b) illustrates an example in which two channels

(channels 1 and 2) configure one channel pair, and block

switching is performed synchronously in channels 1 and 2. Interleaving is also applied in this example. FIG. 6 (c)

illustrates an example in which two channels (channels 1 and

2) configure one channel pair, and the block switching of

channels 1 and 2 is performed independently. Herein, the

channel pair refers to two arbitrary audio channels. The decision on which channels are grouped into channel pairs can

be made automatically by the encoder or manually by the user,

(e.g., L and R channels, Ls and Rs channels) .

In independent block switching, while the length of each

channel may be identical for all channels, the block switching

can be performed individually for each channel. Namely, as

shown in FIG. 6 (c) , the channels may be divided into blocks

differently. If the two channels of a channel pair are

correlated with each other and difference coding is used, both

channels of a channel pair may be block switched synchronously.

In synchronous block switching, the channels are block switched (i.e., divided into blocks) in the same manner. FIG.

6(b) illustrates an example of this, and further illustrates

that the blocks may be interleaved. If the two channels of a

channel pair are not correlated with each other, difference

coding may not provide a benefit, and thus there will be no

need to block switch the channels synchronously. Instead, it

may be more appropriate to switch the channels independently. Furthermore, according to another embodiment of the present

invention, the described method of independent or synchronous block switching may be applied to a multi-channel group having

a number of channels equal to or more than 3 channels . For

example, if all channels of a multi-channel group are

correlated with each other, all channels of a multi-channel

group may be switched synchronously. On the other hand, if all

channels of a multi-channel group are not correlated with each

other, each channel of the multi-channel group may be switched independently.

Moreover, the "bs_info" field is used as the information for

indicating the block switching result. Additionally, the

"bs_info" field is also used as the information for indicating

whether block switching has been performed independently or

performed synchronously for each channel configuring the

channel pair. In this case, as described above, a particular bit (e.g., first bit) within the "bs__info" field may be used.

If, for example, the two channels of the channel pair are independent from one another, the first bit of the ^wbs_info" field is set to "1" . On the other hand, if the two channels

of the channel pair are synchronous to one another, the first

bit of the "bs_info" field is set as "0" .

Hereinafter, FIGs. 6 (a) , 6Cb), and 6(c) will now be described

in detail.

Referring to FIG. 6 (a) , since none of the channels perform

block switching, the related "bs_info" is not generated.

Referring to FIG. 6 (b) , channels 1 and 2 configure a channel

pair, wherein the two channels are synchronous to one another,

and wherein block switching is performed synchronously. For

example, in FIG. 6 (b) , both channels 1 and 2 are split into

blocks of length N/4, both having the same bs__info "bs_info =

(J))IOl 0000". Therefore, one ^λλbs_info" may be transmitted for

each channel pair, which results in reducing the bit rate.

Furthermore, if the channel pair is synchronous, each block

within the channel pair may be required to be interleaved with

one another. The interleaving may be beneficial (or

advantageous). For example, a block of one channel (e.g.,

block 1.2 in FIG. 6 (b) ) within a channel pair may depend on

previous blocks from both channels (e.g., blocks 1.1 and 2.1

in FIG. 6 (b) ) , and so these previous blocks should be available prior to the current one .

Referring to FIG. 6(c), channels 1 and 2 configure a channel

pair. However, in this example, block switching is performed independently. More specifically, channel 1 is split into

blocks of a size (or length) of up to N/4 and has a bs_info of ^Λλbs_info = (1)101 0000". Channel 2 is split into blocks of a size of up to N/2 and has a bs_info of "bs_info = (O₁)IOO 0000".

In the example shown in FIG. 6 (c) , block switching is

performed independently among each channel, and therefore, the

interleaving process between the blocks is not performed. In

other words, for the channel having the blocks switched

independently, channel data may be arranged separately.

Joint Channel Coding

Joint channel coding, also called joint stereo, can be used to

exploit dependencies between two channels of a stereo signal,

or between any two channels of a multi-channel signal. While

it is straightforward to process two channels x_λ(n) and x₂(«)

independently, a simple method of exploiting dependencies between the channels is to encode the difference signal:

d{ή) = x₂{n)-x_λ(ή)

instead of xl (n) or x2 (n) . Switching between X₁ (ή) , x₂(ή)

and d(ή) in each block may be carried out by comparison of

the individual signals, depending on which two signals

can be coded most efficiently. Such prediction with switched difference coding is advantageous in cases where

two channels are very similar to one another. In case of

multi-channel material, the channels can be rearranged by

the encoder in order to assign suitable channel pairs.

Besides simple difference coding, lossless audio codec also supports a more complex scheme for exploiting inter-channel

redundancy between arbitrary channels of multi-channel signals.

Random Access

The present invention relates to audio lossless coding and is able to supports random access. Random access stands for fast

access to any part of the encoded audio signal without costly

decoding of previous parts . It is an important feature for

applications that employ seeking, editing, or streaming of the compressed data. In order to enable random access, within a

random access unit, the encoder needs to insert a frame that

can be decoded without decoding previous frames . The inserted

frame is referred to as a "random access frame" . In such a

random access frame, no samples from previous frames may be used for prediction.

Hereinafter, the information for random access according to

the present invention will be described in detail. Referring

to the configuration syntax (shown in Table 6) , information

related with random access are transmitted as configuration

information. For example, a ^λΛrandom_access" field is used as information for indicating whether random access is allowed,

which may be represented by using 8 bits. Furthermore, if

random access is allowed, the 8-bit "random_access" field designates the number of frames configuring a random access

unit. For example, when "random_access = 0000 0000", the

random access is not supported. In other words, when

"random_access > 0", random access is supported. More

specifically, when "random_access = 0000 0001", this indicates

that the number of frames configuring the random access unit

is 1. This signifies that random access is allowed in all

frame units. Furthermore, when "random_access = 1111 1111",

this indicates that the number of frames configuring the

random access unit is 255. Accordingly, the ^λλrandom_access"

information corresponds to a distance between a random access frame within the current random access unit and a random

access frame within the next random access unit. Herein, the

distance is expressed by the number of frames .

A 32-bit ^ura_unit_size" field is included in the bitstream and

transmitted. Herein, the "ra_unit_size" field indicates the

size from the current random access frame to the next random

access frame in byte units. Accordingly, the "ra_unit_size"

field is either included in the configuration syntax (Table 6)

or included in the frame-data syntax (Table 7) . The

configuration syntax (Table 6) may further include information

indicating a location where the "ra_unit_size" information is stored within the bitstream. This information is represented

as a 2-bit "ra_flag" field. More specifically, for example, when "ra_flag = 00" , this indicates that the "ra_unit_size"

information is not stored in the bitstream. When the "ra_flag

= 01", this indicated that the ^λλra__unit_size" information is stored in the frame-data syntax (Table 7) within the bitstream.

Furthermore, when the "ra_flag = 10", the "ra__unit_size"

information is stored in the configuration syntax (Table 6)

within the bitstream. If the ^Λλra__unit_size" information is

included in the configuration syntax, this indicates that the "ra_unit_size" information is transmitted on the bitstream

only one time and is applied equally to all random access

units. Alternatively, if the "ra_unit_size" information is

included in the frame-data syntax, this indicates the distance

between the random access frame within the current random

access unit and the random access frame within the next random

access unit. Therefore, the "ra_unit_size" information is

transmitted for each random access unit within the bitstream.

Accordingly, the ^wrandom_access" field within the

configuration syntax (Table 6) may also be referred to as

first general information. And, the "ra_flag" field may also

be referred to as second general information. In this aspect

of the present invention, an audio signal includes

configuration information and a plurality of random access

units, each random access unit containing one or more audio data frames, one of which is a random access frame, wherein

the configuration information includes first general

information indicating a distance between two adjacent random

access frames in frames, and second general information

indicating where random access unit size information for each

random access unit is stored. The random access unit size information indicating a distance between two adjacent random

access frames in bytes.

Alternatively, in this aspect of the present invention, a

method of decoding an audio signal includes receiving the

audio signal having configuration information and a plurality

of random access units, each random access unit containing one

or more audio data frames, one of which is a random access

frame, reading first general information from the

configuration information, the first general information

indicating a distance between two adjacent random access

frames in frames, and reading second general information from

the configuration information, the second general information

indicating where random access size information for each

random access unit is stored, and the random access unit size

information indicating a distance between two adjacent random

access frames in bytes.

Channel configuration

As shown in FIG. 3, an audio signal includes multi-channels information according to the present invention. For example, each channel may be mapped at a one-to-one correspondence with

a location of an audio speaker. The configuration syntax

(Table 6 below) includes channel configuration information,

which is indicated as a 16-bit "chan_config__info" field and a

16-bit "channels" field. The "chan_config_info" field includes

information for mapping the channels to the loudspeaker

locations and the 16-bit "channels" field includes information

indicating the total number of channels. For example, when the

"channels" field is equal to "0" , this indicates that the

channel corresponds to a mono channel. When the "channels"

field is equal to "1" , this indicates that the channel

corresponds to one of stereo channels. And, when the

"channels" field is equal to or more than "2", this indicates that the channel corresponds to one of multi-channels .

Table 2 below shows examples of each bit configuring the

"chan_config_info" field and each respective channel

corresponding thereto. More specifically, when a

corresponding channel exists within the transmitted bitstream,

the corresponding bit within the "chan_config_info" field is

set to "1" . Alternatively, when a corresponding channel does

not exist within the transmitted bitstream, the corresponding

bit within the "chan_config__info" field is set to "0" . The

present invention also includes information indicating whether

the "chan_config_info" field exists within the configuration syntax (Table 6) . This information is represented as a 1-bit

"chan_config" flag. More specifically, "chan_config = 0"

indicates that the ^λVchan_config_info" field does not exist.

And, ^wchan_config = 1" indicates that the ^uchan_config_info"

field exists. Therefore, when ^wchan_config = 0", this

indicates that the ^uchan_config_info" field is not newly

defined within the configuration syntax (Table 6) .

Table 2: Channel configuration.

Frame length

As shown in FIG. 3, an audio signal includes multiple or multi-channels according to the present invention. Therefore, when performing encoding, information on the number of multichannels configuring one frame and information on the number of samples for each channel are inserted in the bitstream and transmitted. Referring to the configuration syntax (Table 6) , a 32-bit "samples" field is used as information indicating the total number of audio data samples configuring each channel. Further, a 16-bit ^wframe_length" field is used as information indicating the number of samples for each channel within the corresponding frame.

Furthermore, a 16-bit value of the "frame_length" field is determined by a value used by the encoder, and is referred to as a user-defined value. In other words, instead of being a fixed value, the user-defined value is arbitrarily determined upon the encoding process.

Therefore, during the decoding process, when the bitstream is received through the demultiplexing part 200 of shown in FIG. 2, the frame number of each channel should first be obtained. This value is obtained according to the algorithm shown below.

frames = samples / frame_length; rest = samples % frame_length; if (rest)

{ frames++; frlen_last = rest ;

} else frlen_last = frame_length;

More specifically, the total number of frames for each channel

is calculated by dividing the total number of samples for each

channel, which is decided by the "samples" field transmitted

through the bitstream, by the number of samples within a frame of each channel, which is decided by the "frame_length" field.

For example, when the total number of samples decided by the

"samples" field is an exact multiple of the number of samples

within each frame, which is decided by the "frame_length"

field, the multiple value becomes the total number of frames. However, if the total number of samples decided by the

"samples" field is not an exact multiple of the number of

samples decided by the "frame_length" field, and a remainder

(or rest) exist, the total number of frames increases by "1"

more than the multiple value. Furthermore, the number of

samples of the last frame (frlen_last) is decided as the

remainder (or rest) . This indicates that only the number of

samples of the last frame is different from its previous frame.

By defining a standardized rule between the encoder and the

decoder, as described above, the encoder may freely decide and

transmit the total number of samples ("samples" field) for each channel and the number of samples ("frame_length" field)

within a frame of each channel. Furthermore, the decoder may

accurately decide, by using the above-described algorithm on the transmitted information, the number of frames for each

channel that is to be used for decoding.

Linear Prediction

In the present invention, linear prediction is applied for the

lossless audio coding. The predictor 160 shown in FIG. 1

includes at least one or more filter coefficients so as to

predict a current sample value from a previous sample value .

Then, the second entropy coding part 180 performs entropy

coding on a residual value corresponding to the difference

between the predicted value and the original value .

Additionally, the predictor coefficient values for each block

that are applied to the predictor 160 are selected as optimum

values from the coefficient estimating part 120. Further, the

predictor coefficient values are entropy coded by the first

entropy coding part 140. The data coded by the first entropy

coding part and the second entropy coding part 180 are

inserted as part of the bitstream by the multiplexing part 190

and then transmitted.

Hereinafter, the method of performing linear prediction according to the present invention will now be described in detail . Prediction with FIR Filters

Linear prediction is used in many applications for speech and

audio signal processing. Hereinafter, an exemplary operation

of the predictor 160 will be described based on Finite Impulse

Response (FIR) filters. However, it is apparent that this example will not limit the scope of the present invention.

The current sample of a time-discrete signal x{ή) can be

approximately predicted from previous samples x(n-k) . The

prediction is given by the following equation.

K x(n)=∑h_k*x(n-k),

4=1

wherein K is the order of the predictor. If the

predicted samples are close to the original samples, the

residual shown below:

e(ή) =x(ή) - x(ή)

has a smaller variance than x(n) itself, hence e(ή) can be

encoded more efficiently.

The procedure of estimating the predictor coefficients from a

segment of input samples, prior to filtering that segment is

referred to as forward adaptation. in this case, the coefficients should be transmitted. On the other hand, if the

coefficients are estimated from previously processed segments

or samples, e.g., from the residual, reference is made to

backward adaptation. The backward adaptation procedure has

the advantage that no transmission of the coefficients is needed, since the data required to estimate the coefficients

is available to the decoder as well.

Forward-adaptive prediction methods with orders around 10 are

widely used in speech coding, and can be employed for lossless

audio coding as well. The maximum order of most forward-

adaptive lossless prediction schemes is still rather small, e.g., K = 32. An exception is the special 1-bit lossless

codec for the Super Audio CD, which uses prediction orders of

up to 128.

On the other hand, backward-adaptive FIR filters with some

hundred coefficients are commonly used in many areas, e.g.,

channel equalization and echo cancellation. Most of these

systems are based on the LMS algorithm or a variation thereof,

which has also been proposed for lossless audio coding. Such

LMS-based coding schemes with high orders are applicable since

the predictor coefficients do not have to be transmitted as

side information, thus their number does not contribute to the

data rate. However, backward-adaptive codecs have the

drawback that the adaptation has to be carried out both in the

encoder and the decoder, making the decoder significantly more complex than in the forward-adaptive case.

Forward-Adaptive Prediction

As an exemplary embodiment of the present invention, forward

adaptive prediction will be given as an example in the description set forth herein. In forward-adaptive linear

prediction, the optimal predictor coefficients h_k (in terms of

a minimized variance of the residual) are usually estimated

for each block by the coefficient estimating part 120 using

the autocorrelation method or the covariance method. The

autocorrelation method, using the conventional Levinson-Durbin algorithm, has the additional advantage of providing a simple

means to iteratively adapt the order of the predictor.

Furthermore, the algorithm inherently calculates the

corresponding parcor coefficients as well.

Another aspect of forward-adaptive prediction is to determine

a suitable prediction order. Increasing the order decreases

the variance of the prediction error, which leads to a smaller

bit rate R_e for the residual. On the other hand, the bit rate

R_c for the predictor coefficients will rise with the number of

coefficients to be transmitted. Thus, the task is to find the

optimum order which minimizes the total bit rate. This can be expressed by minimizing the equation below: , R_tolal(K)=R₆(K)₊R₀(K),

with respect to the prediction order K. As the prediction gain rises monotonically with higher orders, Re decreases with

K. On the other hand R_c rises monotonically with K, since an

increasing number of coefficients should be transmitted.

The search for the optimum order can be carried out

efficiently by the coefficient estimating part 120, which

determines recursively all predictors with increasing order.

For each order, a complete set of predictor coefficients is

calculated. Moreover, the variance σ_e ² of the corresponding

residual can be derived, resulting in an estimate of the

expected bit rate for the residual. Together with the bit rate for the coefficients, the total bit rate can be

determined in each iteration, i.e., for each prediction order.

The optimum order is found at the point where the total bit

rate no longer decreases .

While it is obvious from the above equation that the

coefficient bit rate has a direct effect on the total bit rate,

a slower increase of R_c also allows to shift the minimum of

R_tota] to higher orders (wherein R_e is smaller as well) , which

would lead to better compression. Hence, efficient yet

accurate quantization of the predictor coefficients plays an

important role in achieving maximum compression. Prediction orders

In the present invention, the prediction order K, which

decides the number of predictor coefficients for linear

prediction, is determined. The prediction order K is also determined by the coefficient estimating part 120. Herein,

information on the determined prediction order is included in

the bitstream and then transmitted.

The configuration syntax (Table 6) includes information

related to the prediction order K. For example, a 1-bit to

10-bit "max_order" field corresponds to information indicating

a permitted order value. The highest value of the 1-bit to

10-bit "max_order" field is K=1023 (e.g. , 10-bit) . As another information related to the prediction order K, the

configuration syntax (Table 6) includes a 1-bit "adapt_order"

field, which indicates whether an optimum order for each block

exists. For example, when ^wadapt_order = 1", an optimum order

should be provided for each block. In a block_data syntax

(Table 8) , the optimum order is provided as a 1-bit to 10-bit

"opt_order" field. Further, when ^λΛadapt_order = 0", a

separate optimum order is not provided for each block. In

this case, the "max_order" field becomes the final order applied to all of the blocks .

The optimum order (opt_order) is decided based upon the value

of max_order field and the size (N_B) of the corresponding block. More specifically, for example, when the max_order is

decided as K_maχ = 10 and ^λΛadapt_order = 1" , the opt_order for each block may be decided considering the size of the

corresponding block. In some case, the opt_order value being

larger than max_order (K_max = 10) is possible.

In particular, the present invention relates to higher

prediction orders . In the absence of hierarchical block

switching, there may be a factor of 4 between the long and the

short block length (e.g. 4096 S- 1024 or 8192 & 2048), in

accordance with the embodiments. On the other hand, in the

embodiments where hierarchical block switching is implemented,

this factor can be increased (e.g., up to 32), enabling a

larger range (e.g., 16384 down to 512 or even 32768 to 1024

for high sampling rates) .

In the embodiments where hierarchical block switching is

implemented, in order to make better use of very long blocks,

higher maximum prediction orders may be employed. The maximum

order may be K_msκ= 1023. In the embodiments, ϋ_^" _max may be bound

by the block length N_B, for example, Z_max < N_B / 8 (e.g., K_msκ =

255 for N_B = 2048). Therefore, using K_max = 1023 may require a

block length of at least N_B = 8192. In the embodiments, the

^wmax_order" field in the configuration syntax (Table 6) can be

up to 10 bits and "opt_order" field in the block_data syntax

(Table 8) can also be up to 10 bits. The actual number of bits in a particular block may depend on the maximum order

allowed for a block. If the block is short, a local

prediction order may be smaller than a global prediction order.

Herein, the local prediction order is determined from

considering the corresponding block length N_B, and the global

prediction order is determined from the ^λλmax_order" K_max in the

configuration syntax. For example, if K_max = 1023, but N_B =

2048, the "opt_order" field is determined on 8 bits (instead

of 10) due to a local prediction order of 255.

More specifically, the opt_order may be determined based on

the following equation:

opt_order = min (global prediction order, local prediction order) ;

And. the global and local prediction orders may be determined

by:

global prediction order = ceil (Iog2 (permitted prediction

order +1) )

local prediction order = max (ceil (Iog2 ( (Nb>>3) -1) ), 1)

In the embodiments, data samples of the subdivided block from

a channel are predicted. A first sample of a current block is predicted using the last K samples of a previous block. The K value is determined from the opt_order which is derived from the above-described equation.

If the current block is a first block of the channel, no samples from the previous block are used. In this case, prediction with progressive order is employed. For example, assuming that the opt_order value is K=5 for a corresponding block, the first sample in the block does not perform prediction. The second sample of the block uses the first sample of the block to perform the prediction (as like K=I) , the third sample of the block uses the first and second samples of the block to perform the prediction (as like K=2) , etc. Therefore, starting from the sixth sample and for samples thereafter, prediction is performed according to the opt_order of K=5. As described above, the prediction order increases progressively from K=I to K=5.

The above-described progressive order type of prediction is very advantageous when used in the random access frame. Since the random access frame corresponds to a reference frame of the random access unit, the random access frame does not perform prediction by using the previous frame sample. Namely, this progressive prediction technique may be applied at the beginning of the random access frame.

Quantization of Predictor Coefficients The above-described predictor coefficients are quantized in the quantizing part 130 of FIG. 1. Direct quantization of the

predictor coefficients h_k is not very efficient for

transmission, since even small quantization errors may result

in large deviations from the desired spectral characteristics of the optimum prediction filter. For this reason, the

quantization of predictor coefficients is based on the parcor

(reflection) coefficients r_k , which can be calculated by the

coefficient estimating part 120. As described above, for example, the coefficient estimating part 120 is processed

using the conventional Levinson-Durbin algorithm.

The first two parcor coefficients ( ^{^1} and '² correspondingly) are quantized by using the following functions:

while the remaining coefficients are quantized using simple 7- bit uniform quantizers:

In all cases the resulting quantized values a_k are restricted

to the range [-64,63] .

Entropy Coding As shown in FIG. 1, two types of entropy coding are applied in the present invention. More specifically, the first entropy coding part 140 is used for coding the above-described predictor coefficients. And, the second entropy coding part 180 is used for coding the above-described audio original samples and audio residual samples. Hereinafter, the two types of entropy coding will now be described in detail.

First Entropy Coding of the predictor coefficient The related art Rice code is used as the first entropy coding method according to the present invention. For example,

transmission of the quantized coefficients a_k is performed by

producing residual values:

δ_k = a_k -offset_A ,

which, in turn, are encoded by using the first entropy coding part 140, e.g., the Rice code method. The corresponding offsets and parameters of Rice code used in this process can be globally chosen from one of the sets shown in Table 3, 4 and 5 below. A table index (i.e., a 2-bit "coef_table" ) is indicated in the configuration syntax (Table 6) . If "coef_table = 11", this indicates that no entropy coding is applied, and the quantized coefficients are transmitted with 7 bits each. In this case, the offset is always -64 in order to

obtain unsigned values δ_k=a_k+64 that are restricted to

[0,127]. Conversely, if "coeff_table = 00", Table 3 below is selected, and if "coeff_table = 01", Table 4 below is selected.

Finally, if "coeff_table = 11", Table 5 is selected.

When receiving the quantized coefficients in the decoder of

FIG.2, the first entropy decoding part 220 reconstructs the

predictor coefficients by using the process that the residual

values δ_k axe. combined with offsets to produce quantized

indices of parcor coefficients α_k :

α_k = δ_k+oϋsQt_k

Thereafter, the reconstruction of the first two coefficients

[γ_γ and γ₂) is performed by using:

wherein 2^e represents a constant (β=20) scale factor required

for integer representation of the reconstructed coefficients,

and rQ is an empirically determined mapping table (not shown

as the mapping table may vary with implementation) .

Accordingly, the three types of coefficient tables used for

the first entropy coding are provided according to the sampling frequency. For example, the sampling frequency may be divided to 48kHz, 96kHz, and 192kHz. Herein, each of the

three Tables 3, 4, and 5 is respectively provided for each

sampling frequency. Instead of using a single table, one of three different tables can be chosen for the entire file. The table should

typically be chosen depending on the sampling rate. For

material with 44.1 kHz, the applicant of the present invention

recommends to use the 48 kHz table. However, in general, the

table can also be chosen by other criteria.

Table 3 : Rice code parameters used for encoding of

quantized coefficients (48 kHz) .

Table 4 : Rice code parameters used for encoding of

quantized coefficients (96 kHz) .

Table 5 : Rice code parameters used for encoding of

quantized coefficients (192 kHz) .

Second Entropy Coding of the Residual

The present invention contains two different modes of the coding method applied to the second entropy coding part 180 of

FIG. 1, which will now be described in detail.

In the simple mode, the residual values e(n) are entropy coded

using Rice code. For each block, either all values can be

encoded using the same Rice code, or the block can be further

divided into four parts, each encoded with a different Rice

code. The indices of the applied codes are transmitted, as

shown in FIG. 1. Since there are different ways to determine

the optimal Rice code for a given set of data, it is up to the

encoder to select suitable codes depending upon the statistics of the residual.

Alternatively, the encoder can use a more complex and

efficient coding scheme using BGMC mode. In the BGMC mode,

the encoding of residuals is accomplished by splitting the distribution in two categories . The two types include

residuals that belong to a central region of the distribution,

e(n)|<e_max , and residuals that belong to its tails. The

residuals in tails are simply re-centered (i.e., for e(n) > e_max ,

e_t(n) = e(ή) -e_max is provided) and encoded using Rice code as

described above. However, in order to encode residuals in the

center of the distribution, the BGMC first splits the

residuals into LSB and MSB components, then the BGMC encodes MSBs using block Gilbert-Moore (arithmetic) codes. And

finally, the BGMC transmits LSBs using direct fixed-lengths

codes . Both parameters e _max and the number of directly

transmitted LSBs may be selected such that they only slightly

affect the coding efficiency of this scheme, while allowing the coding to be significantly less complex.

The configuration syntax (Table 6) and the block_data syntax

(Table 8) according to the present invention include

information related to coding of the Rice code and BGMC code. The information will now be described in detail

The configuration syntax (Table 6) first includes a 1-bit

^λΛbgmc_mode" field. For example, ^λxbgmc_mode = 0" signifies the

Rice code, and ^λλbgmc_mode = 1" signifies the BGMC code. The

configuration syntax (Table 6) also includes a 1-bit "sb_j?art"

field. The "sb_part" field corresponds to information related

to a method of partitioning a block to a sub-block and coding the partitioned sub-block. Herein, the meaning of the

"sb__part" field varies in accordance with the value of the

^λλbgmc_mode" field.

For example, when "bgmc_mode = 0", in other words when the

Rice code is applied, "sb_j?art = 0" signifies that the block

is not partitioned into sub-blocks. Alternatively, ^wsb_part =

1" signifies that the block is partitioned at a 1:4 sub-block partition ratio. Additionally, when "bgmc_mode = 1", in other

words when the BGMC code is applied, ^Λλsb_part = 0" signifies

that the block is partitioned at a 1:4 sub-block partition

ratio. Alternatively, ^λλsb_jpart = 1" signifies that the block

is partitioned at a 1:2:4:8 sub-block partition ratio. The block_data syntax (Table 8) for each block corresponding

to the information included in the configuration syntax (Table

6) includes 0-bit to 2-bit variable "ec_sub" fields. More

specifically, the "ec_sub" field indicates the number of sub-

blocks existing in the actual corresponding block. Herein,

the meaning of the "ec_sub" field varies in accordance with

the value of the "bgmcjnode" + "sb_part" fields within the

configuration syntax (Table S) .

For example, ^λXbgmc_mode + sb_part = 0" signifies that the Rice

code does not configure the sub-block. Herein, the "ec_sub"

field is a 0-bit field, which signifies that no information is included.

In addition, ^Λλbgmc_mode + sbjpart = 1" signifies that the Rice code or the BGMC code is used to partition the block to sub-

blocks at a 1:4 rate. Herein, only 1 bit is assigned to the

^λXec_sub" field. For example, "ec_sub = 0" indicates one sub-

block (i.e., the block is not partitioned to sub-blocks), and

"ec_sub = 1" indicates that 4 sub-blocks are configured.

Furthermore, ^Λλbgmc_mode + sb_jpart = 2" signifies that the BGMC

code is used to partition the block to sub-blocks at a 1:2:4:8

rate. Herein, 2 bits are assigned to the ^wec_sub" field. For

example, "ec_sub = 00" indicates one sub-block (i.e., the

block is not partitioned to sub-blocks) , and, "ec_sub = 01"

indicates 2 sub-blocks. Also, ^Λλec_sub = 10" indicates 4 sub-

blocks, and "ec_sub = 11" indicates 8 sub-blocks.

The sub-blocks defined within each block as described above

are coded by second entropy coding part 180 using a difference coding method. An example of using the Rice code will now be

described. For each block of residual values, either all

values can be encoded using the same Rice code, or, if the

"sb_part" field in the configuration syntax is set, the block

can be partitioned into 4 sub-blocks, each encoded sub-block

having a different Rice code. In the latter case, the

"ec_sub" field in the block-data syntax (Table 8) indicates

whether one or four blocks are used.

While the parameter s [i = 0] of the first sub-block is

directly transmitted with either 4 bits (resolution < 16 bits)

or 5 bits (resolution > 16 bits), only the differences (s[i]~ s[i-l]) of following parameters s [i > 0] are transmitted.

These differences are additionally encoded using appropriately

chosen Rice codes again. In this case, the Rice code parameter

used for differences has the value of "0".

Syntax

According to the embodiment of the present invention, the

syntax of the various information included in the audio

bitstream are shown in the tables below. Table 6 shows a

configuration syntax for audio lossless coding. The

configuration syntax may form a header periodically placed in

the bitstream, may form a header of each frame; etc. Table 7

shows a frame-data syntax, and Table 8 shows a block-data syntax.

Table 6: Configuration syntax.

Table 7: Frame data syntax.

Syntax Bits frame_data ( )

{ if ( (ra_flag == 1) && (frame_id % random_access == 0) ) { ra_unit_size 32

} if (mc_coding && j oint_stereo) { js_switch; byte align; } if ( !mc_coding | | j s_switch) { for (c = 0 ; c < channels ; C++) { if (block_switching) { bs_info; 8,16_# 32

} if (independent_bs) { for (b = 0; b < blocks; b++) { block_data(c) ;

} else{ for (b = 0; b < blocks; b++) { block_data (c) ; block_data(c+l) ;

}

C++;

} } else{ if (block_switching) { bs_info; 8,16, 32

} for (b = 0; b < blocks; b++) { for (c = 0; c < channels; C++) { block_data (c) ; channe1 data (c) ;

}

}

} if (floating)

{ num_bytes_diff_float; 32 diff_float_data ( ) ; }

Table 8: Block data syntax.

Syntax Bits block_data ( )

{ block_type; if (block_type == 0) { const_block; js block; (reserved) if (const block == 1) if (resolution == 8) { const_val;

} else if (resolution == 16) { const_val; 16

} else if (resolution == 24) { const_val; 24

} else { const_val; 32

}

}

} else { js_block; if ( (bgmc_mode == 0) && (sb_part == 0) { sub_blocks = 1;

} else if ( (bgmc_mode == 1) && (sb_j>art { ec_sub; sub_blocks = 1 << ec_sub;

} else { ec_sub; sub_blocks = (ec_sub == 1) ? 4 : 1;

} if (bgmc_mode == 0) { for (k = 0; k < sub_blocks; k++) { S [k] ; varxe s

} else { for (k = 0 ; k < sub_blocks ; k++) { s [k] , sx [k] ; varxe s

} sb_length = block_length / sub_blocks; shift_lsbs; 1 if (shift_lsbs == 1) { shift_pos; 4

} if (!RLSLMS) { if (adapt_order == 1) { opt order? 1...10

Compression Results

In the following, the lossless audio codec is compared with

two of the most popular programs for lossless audio compression: the open-source codec FLAC and the Monkey's Audio

(MAC 3.97). Herein, the open-source codec FLAG uses forward-

adaptive prediction, and the Monkey's Audio (MAC 3.97) is a

backward-adaptive codec used as the current state-of-the-art

algorithm in terms of compression. Both codecs were run with

options providing maximum compression (i.e., flac -8 and mac-

c4000) . The results for the encoder are determined for a

medium compression level (with the prediction order restricted

to K _ 60) and a maximum compression level (K _ 1023) , both

with random access of 500 ms . The tests were conducted on a 1.7 GHz Pentium-M system, with 1024 MB of memory. The test

comprises nearly 1 GB of stereo waveform data with sampling

rates of 48, 96, and 192 kHz, and resolutions of 16 and 24

bits.

Compression Ratio

In the following, the compression ratio is defined as: _{c =} CompressedFileSize * _{l 0QO/o} ^ OriginalFileSize

wherein smaller values indicate better compression. The

results for the examined audio formats are shown in Table 9

(192 kHz material is not supported by the FLAG codec) .

Table 9: Comparison of average compression ratios for

different audio formats (kHz/bits) .

The results show that ALS at maximum level outperforms both

FLAC and Monkey's Audio for all formats, but particularly for

high-definition material (i.e., 96 kHz / 24-bit and above).

Even at medium level, ALS delivers the best overall

compression.

Complexity

The complexity of different codecs strongly depends on the

actual implementation, particularly that of the encoder. As mentioned above, the audio signal encoder of the present

invention is an ongoing development. Thus, we restrict our

analysis to the decoder, a simple C code implementation with

no further optimizations. The compressed data were generated

by the currently best encoder implementation. The average CPU

load for real-time decoding of various audio formats, encoded at different complexity levels, is shown in Table 10. Even

for maximum complexity, the CPU load of the decoder is only

around 20-25%, which in return means that file based decoding

is at least 4 to 5 times faster than real-time.

Table 10 : Average CPU load (percentage on a 1.7 GHz

Pentium-M) , depending on audio format (kHz/bits) and ALS

encoder complexity.

The codec is designed to offer a large range of complexity

levels. While the maximum level achieves the highest

compression at the expense of slowest encoding and decoding

speed, the faster medium level only slightly degrades

compression, but decoding is significantly less complex than

for the maximum level (i.e., approximately 5% CPU load for 48 kHz material). Using a low-complexity level (i.e., K _ 15,

Rice coding) degrades compression by only 1 to 1.5% compared to the medium level, but the decoder complexity is further

reduced by a factor of three (i.e., less than 2% CPU load for

48 kHz material) . Thus, audio data can be decoded even on hardware with very low computing power.

While the encoder complexity may be increased by both higher

maximum orders and a more elaborate block switching algorithm

(in accordance with the embodiments) , the decoder may be

affected by a higher average prediction order.

The foregoing embodiments (e.g., hierarchical block switching)

and advantages are merely examples and are not to be construed

as limiting the appended claims. The above teachings can be

applied to other apparatuses and methods, as would be appreciated by one of ordinary skill in the art. Many

alternatives, modifications, and variations will be apparent to those skilled in the art.

Industrial Applicability

It will, be apparent to those skilled in the art that various

modifications and variations can be made in the present

invention without departing from the spirit or scope of the

inventions. For example, aspects and embodiments of the

present invention can be readily adopted in another audio

signal codec like the lossy audio signal codec. Thus, it is intended that the present invention covers the modifications and variations of this invention.

Claims

[CLAIMS]

1. A method of processing an audio signal, comprising: subdividing a channel in a frame of the audio signal into a plurality of blocks, at least two of the blocks having different lengths; and determining an optimum prediction order for each block based on a permitted prediction order and a length of the block.

2. The method of claim 1, wherein the determining step comprises : determining a global prediction order based on the permitted prediction order; determining a local prediction order based on the length of the block; and selecting a minimum one of the global prediction order and the local prediction order as the optimum prediction order.

3. The method of claim 2, wherein the determining a global prediction order step determines the global prediction order as equal to,

ceil (Iog2 (permitted prediction order +1) .

4. The method of claim 2, wherein the determining a local prediction order step determines the local prediction order as

equal to,

max (ceil (Iog2 ( (Nb>>3) -1) ) , 1)

where Nb is the length of the block.

5. The method of claim 2, wherein

the determining a global prediction order step determines the global prediction order as equal to,

ceil (Iog2 (permitted prediction order +1); and

the determining a local prediction order step determines the local prediction order as equal to,

max(ceil(log2( (Nb>>3) -1) ) , 1)

where Nb is the length of the block.

6. The method of claim 1, further comprising: predicting current data samples in the channel based on previous data samples, a number of the previous data samples used in the predicting step being equal to the optimum

prediction order; and obtaining a residual of the current data samples based

on the predicted data samples .

7. The method of claim 6, wherein the predicting step

progressively increases the prediction order to the optimum

prediction order as previous data samples become available.

8. The method of claim 7, wherein

the prediction order for an initial data sample having no

previous data samples is zero, and the predicting step

generates zero as the predicted initial data sample; and

the obtaining step effectively does not change the

initial data sample as a result of the predicted initial

data sample being zero.

9. The method of claim 6, wherein the determining step

comprises :

determining a global prediction order based on the

permitted prediction order;

determining a local prediction order based on the length

of the block; and

selecting a minimum one of the global prediction order

and the local prediction order as the optimum prediction order.

10. The method of claim 9, wherein

the determining a global prediction order step determines

the global prediction order as equal to,

ceil (Iog2 (permitted prediction order +1); and

the determining a local prediction order step determines

the local prediction order as equal to,

max (ceil (Iog2 ( (Nb>>3) -1) ) , 1)

where Nb is the length of the block.

11. The method of claim 1, wherein the subdividing step

subdivides the channel according to a subdivision hierarchy,

the subdivision hierarchy has more than one level, and each

level is associated with a different block length.

12. The method of claim 11, wherein a superordinate level of

the subdivision hierarchy is associated with a block length

double a block length associated with a subordinate level.

13. The method of claim 11, wherein the subdividing step selectively subdivides a block of the super ordinate level to

obtain two blocks at the subordinate level.

14. The method of claim 11, wherein if a channel has a length

of N, then the subdividing step subdivides the channel into the plurality of blocks such that each block has a length of

one of N/2, N/4, N/8, N/16 and N/32.

15. The method of claim 11, wherein the subdividing step

subdivides the channel into the plurality of blocks such that

each block has a length equal to one of ,

N/ (m¹) for i = 1, 2, ... p,

where N is the length of the channel, m is an integer greater

than or equal to 2, and p represents a number of the levels in

the subdivision hierarchy.

16. The method of claim 15, wherein m = 2 and p = 5.

17. The method of claim 11, further comprising:

generating information indicating the subdivision of the

channel into the blocks .

18. The method of claim 17, wherein the generating step generates the information such that a length of the

information depends on a number of levels in the subdivision

hierarchy.

19. The method of claim 17, wherein the generating step generates the information such that the information includes a

number of information bits, and the information bits indicate

the subdivision of the channel into the blocks .

20. The method of claim 19, wherein each information bit is

associated with a level in the subdivision hierarchy and is

associated with a block at the associated level, and each

information bit indicates whether the associated block was subdivided,

21. The method of claim 20, wherein if the information bit has

a value of 1, then the associated block is subdivided, and if

the information bit has a value of 0, then the associated

block is not subdivided.

22. The method of claim 17, further comprising:

sending the information.

23. The method of claim 11, wherein if the frame is a random

access frame, which is a frame to be encoded such that previous frames are not necessary to decode the frame,

the predicting step progressively increases the

prediction order to a desired prediction order as

previous data samples from only the random access channel

become available.

24. A method of encoding an audio signal, comprising:

subdividing a channel in a frame of the audio signal

into a plurality of blocks, at least two of the blocks having

different lengths; and

determining an optimum prediction order for each block

based on a permitted prediction order and a length of the

block; and

encoding the plurality of blocks to produce a compressed bitstream based on the optimum prediction order.

25. A method of decoding an audio signal, comprising:

receiving an audio data frame having at least one

channel, the channel being subdivided into a plurality of

blocks, at least two of the blocks having different lengths,-

obtaining information from the audio signal indicating

the optimum prediction order for each block; and

decoding the channel based on the obtained information.

26. An apparatus for encoding an audio signal, comprising: an encoder configured to subdivide at least one channel

in a frame of the audio signal into a plurality of blocks, at

least two of the blocks having different lengths, and the

encoder configured to determine an optimum prediction order

for each block based on a permitted prediction order and a

length of the block and encode the plurality of blocks to

produce a compressed bitstream based on the optimum prediction

order .

27. An apparatus for decoding an audio signal, comprising:

a decoder configured to receive an audio data frame

having at least one channel, the channel being subdivided into

a plurality of blocks, at least two of the blocks having

different lengths, and the decoder configured to obtain

information from the audio signal indicating the optimum

prediction order for each block and decode the channel based on the obtained information.