FORM 2
THE PATENTS ACT 1970
[39 OF 1970]
COMPLETE SPECIFICATION
[See Section 10; rule 13]
"A PROGRESSIVE TO LOSSLESS EMBEDDED AUDIO CODER (PLEAC) WITH MULTIPLE FACTORIZATION REVERSIBLE TRANSFORM"
MICROSOFT CORPORATION, a corporation of the State of Washington having a place of business at One Microsoft Way, Redmond, Washington 98052, United States of America,
The following specification particularly describes the nature of the invention and the manner in which it is to be performed:-


MS303116.1

Express NL Ao. ET800619983US

Title: A Progressive to Lossless Embedded Audio Coder (PLEAC) with Multiple Factorization Reversible Transform
TECHNICAL FIELD
5 The present invention relates generally to data coding, and more particularly to a
system and method facilitating audio and/or picture encoding and/or decoding.
BACKGROUND OF THE INVENTION The amount of information available via computers has dramatically increased
10 with the wide spread proliferation of computer networks, the Internet and digital storage
means. With such increased amount of information has come the need to transmit information quickly and to store the information efficiently. Data compression is a technology that facilitates the effective transmitting and storing of information
Data compression reduces an amount of space necessary to represent information,
15 and can be used for many information types. The demand for compression of digital
information, including images, text, audio and video has been ever increasing. Typically, data compression is used with standard computer systems; however, other technologies make use of data compression, such as but not limited to digital and satellite television as well as cellular/digital phones.
20 As the demand for handling, transmitting and processing large amounts of
information increases, the demand for compression of such data increases as well. Although storage device capacity has increased significantly, the demand for information has outpaced capacity advancements. For example, an uncompressed digital music source can require 5 megabytes of space whereas the same music can be compressed
25 without loss and require only 2.5 megabytes of space. Thus, data compression facilitates
transferring larger amounts of information. Even with the increase of transmission rates, such as broadband, DSL, cable modem Internet and the like, transmission limits are easily reached with uncompressed information. For example, transmission of an uncompressed music over a DSL line can take ten minutes. However, the same music

MS303116.1
can be transmitted in about one minute when compressed thus providing a ten-fold gain in data throughput.
In general, there are two types of compression, lossless and lossy. Lossless compression allows exact original data to be recovered after compression, while lossy
5 compression allows for data recovered after compression to differ from the original data.
A tradeoff exists between the two compression modes in that lossy compression provides for a better compression ratio than lossless compression because some degree of data integrity compromise is tolerated. Lossless compression may be used, for example, when compressing critical audio recording, because failure to reconstruct exactly die data can
10 dramatically affect quality and analysis of the audio content. Lossy compression can be
used with consumer music or non-critical audio recording where a certain amount of distortion or noise is tolerable to human senses.
Audio compression is an important technical problem. Most Web pages today host digital music, and digital music playing devices have become increasing popular
15 these days.
Further, there are many existing schemes for encoding audio files. Several such schemes attempt to achieve higher compression rations by using known human psychoacoustic characteristics to mask the audio file. A psychoacoustic coder is an audio encoder which has been designed to take advantage of human auditory masking by
20 dividing the audio spectrum of one or more audio channels into narrow frequency bands
of different sizes optimized with respect to the frequency selectivity of human hearing. This makes it possible to sharply filter coding noise so that it is forced to stay very close in frequency to the frequency components of the audio signal being coded. By reducing the level of coding noise wherever there are no audio signals to mask it, the sound quality
25 of the original signal can be subjectively preserved.
In fact, virtually all state-of-the-art audio coders, including the G.722.1 coder, the MPEG-1 Layer 3 coder, the MPEG-2 AAC coder, and the MPEG-4 T/F coder, recognize the importance of the psychoacoustic characteristics, and adopt auditory masking techniques in coding audio files. In particular, using human psychoacoustic hearing
30 characteristics in audio file compression allows for fewer bits to be used to encode audio
components that are less audible to the human ear. Conversely, more bits can then be
2

MS303116.1
used to encode any psychoacoustic components of the audio file to which the human ear is more sensitive. Such psychoacoustic coding makes it possible to greatly improve the quality of an encoded audio at given bit rate.
Psychoacoustic characteristics are typically incorporated into an audio coding
5 scheme in the following way. First, the encoder explicitly computes auditory masking
thresholds of a group of audio coefficients, usually a "critical band," to generate an "audio mask." These thresholds are then transmitted to the decoder in certain forms, such as, for example, the quantization step size of the coefficients. Next, the encoder quantizes the audio coefficients according to the auditory mask. For auditory sensitive
10 coefficients, i.e., those to which the human ear is more sensitive, a smaller quantization
step size is typically used. For auditory insensitive coefficients, i.e., those to which the human ear is less sensitive, a larger quantization step size is typically used. The quantized audio coefficients are then typically entropy encoded, either through a Huffman coder such as the MPEG-4 AAC quantization & coding, a vector quantizer such
15 as the MPEG-4 TwinVQ, or a scalable bitplane coder such as the MPEG-4 BSAC coder.
In each of the aforementioned conventional audio coding schemes, the auditory masking is applied before the process of entropy coding. Consequently, the masking threshold is transmitted to the decoder as overhead information. As a result, the quality of the encoded audio at a given bit rate is reduced to the extent of the bits required to
20 encode the auditory masking threshold information.
High performance audio codec brings digital music into reality. Popular audio compression technologies, such as MP3, MPEG-4 audio, Real™ and Windows Media Audio (WMA™), are usually lossy in nature. The audio waveform is distorted in exchange for higher compression ratio. In quality critical applications such as a
25 recording/editing studio, it is imperative to maintain the best sound quality possible, i.e.,
the audio should be compressed in a lossless fashion. Since lossless compression ratio is usually limited, it is desirable that the lossless compressed bitstream be scaled to a lossy bitstream of high compression ratio. Most lossless audio coding approaches, simply build upon a lossy audio coder, and further encode the residue. The compression ratio of such
30 approaches is often affected by the underlying lossy coder. Since the quantization noise
in the lossy coder is difficult to model, the approaches usually lead to inefficiency in the
3

MS303116.1
lossless audio coding. Moreover, it is also more complex, as it requires a base coder and a residue coder. Some other approaches build the lossless audio coder directly through a predictive filter and then encode the prediction residue. The approaches may achieve good compression ratio, however, it is not compatible with existing lossy audio coding
5 framework. Since the compression ratio of a lossless coder is rather limited, usually 2-
3:1, the ability to scale a lossless bit stream is very useful. The bit stream generated by the predictive filter based lossless coder cannot be scaled. A lossy/residue coder can generate a bit stream with two layers, a lossy base layer and a lossless enhancement layer. However, the scaling cannot go beyond the lossy base layer. If further scaling in the
10 lossless enhancement layer is required, it is necessary to match the design of the residue
coder with that of the lossy coder, which causes significant complications. Some other approaches build the lossless audio coder directly through a predictive filter and then encode the prediction residue. Though achieving a good compression ratio, such approach is not compatible with existing lossy audio coding framework. Moreover, the
15 resultant bitstream cannot be scaled.
SUMMARY OF THE INVENTION
The following presents a simplified summary of the invention in order to provide
a basic understanding of some aspects of the invention. This summary is not an extensive
20 overview of the invention. It is not intended to identify key/critical elements of the
invention or to delineate the scope of the invention. Its sole purpose is to present some concepts of the invention in a simplified form as a prelude to the more detailed description that is presented later.
The present invention provides for a system and method for lossless and/or
25 progressive to lossless data encoding. The system and method employ a multiple
factorization reversible transform component that provides quantized coefficients that mimic a non-reversible linear transform as closely as possible.
In accordance with an aspect of the present invention, a lossless data coder system
comprising a multiple factorization reversible transform component and an entropy
30 encoder is provided. The system can losslessly encode data (e.g., image, picture and/or
audio). The system receives a digital input (e.g., from an analog to digital converter)
4

MS303116.1
associated with, for example, an audio waveform. The system can be employed, for example, in a vast array of audio and/or document image applications, including, but not limited to, digital audio systems, segmented layered image systems, photocopiers, document scanners, optical character recognition systems, personal digital assistants, fax
5 machines, digital cameras, digital video cameras and/or video games.
Conventionally, a reversible transform component(s) receive an integer input signal and provide an output of quantized coefficients corresponding to the input signal. It is said to be reversible because with such transform, it is possible to exactly recover the input signal from the quantized coefficient. With a reversible transform, lossless
10 compression of the input signal becomes the lossless compression of the quantized
coefficients, which are decorrelated and are easy to compress. Traditionally, a reversible transform is designed with a blueprint from a non-reversible linear transform, and by making substantially every component of the linear transform reversible. However, in such a design process, the quantization noise, which is the difference between the
15 outcome of the reversible transform and its linear transform counterpart, can be large.
In accordance with an aspect of the present invention, in order to reduce quantization noise of the reversible transform, a multiple factorization reversible transform component is used which can factor a linear transform component with multiple reversible forms. In one example, a linear rotation operation is factorized into
20 four different reversible operations. The core of all operations is a three step non-linear
operation. However, the pair of input/output variables can be swapped before and after the certain non-linear operation. The sign of the input/output can be changed as well in certain other cases. The additional forms of factorization lead to different parameters for substantially the same rotation angle, and, thus allows the reversible transform to mimic
25 that of a linear non-reversible transform (e.g., as closely as possible).
The multiple factorization reversible transform component can be employed in an N-point MLT in accordance with one aspect of the present invention. The MLT can comprise a modulation stage, a pre-FFT rotation stage, a complex FFT stage and a post-FFT rotation stage.
30 The modulation stage and pre-FFT rotation stage comprise multiple factorization
rotation operation(s). The core of the complex FFT stage is a multiple factorization
5

MS303116.1
butterfly calculation, which is in turn implemented via multiple factorization rotation. The core of the post-FFT rotation stage is the conjugate rotation operation, which can be implemented by changing the sign of the imaginary part after a normal rotation.
The entropy encoder digitally entropy encodes the quantized coefficients and can
5 employ any suitable entropy encoding technique(s).
Another aspect of the present invention provides for a progressive to lossless data coder system comprising a multiple factorization reversible transform component and an embedded entropy encoder. The progressive to lossless data coder system can be employed to code audio and/or images. The multiple factorization reversible transform
10 component coupled with the embedded entropy encoder (e.g., a high performance
embedded entropy codec) empowers the system with both lossless capability and fine granular scalability progressiveness. The system generates a bit stream that if fully decoded, substantially recovers the original waveform (e.g., audio) without loss and facilitates lossless compression. If a higher compression ratio is desired, the application
15 may extract a subset of the compressed bit stream and form a higher compression ratio bit
stream of lossy nature. Such scaling can be performed in a very large bit rate range, with granularity down to a single byte. With the progressive to lossless functionality of the system, the application can easily balance between the amount of compression required and the desired quality, from a fairly high compression ratio all the way to lossless (e.g.,
20 based on available bandwidth and/or user preference).
The embedded entropy encoder can employ any suitable embedded entropy encoding technique(s). In one example, the embedded entropy encoder employs a fully scalable psychoacoustic audio coder that derives auditory masking thresholds from previously coded coefficients, and uses the derived thresholds for optimizing the order of
25 coding.
Yet another aspect of the present invention provides for a progressive to lossless audio coder system comprising a reversible multiplexer, multiple factorization reversible transform components), embedded entropy encoders) and a bit stream assembler. The system receives an input audio waveform which first goes through the reversible
30 multiplexer. If the input audio is stereo, it is separated into L + R and L - R components.
The waveform of each audio component is then transformed by the multiple factorization
6

MS303116.1
reversible transform component(s) (e.g., a multiple factorization reversible modulated lapped transform).
The reversible multiplexer receives, for example, a stereo audio input. The
reversible multiplexer separates the stereo audio input into L+R and L-R components,
5 where L and R represent the waveform on the left and right audio channel, respectively.
If the input audio is mono, the reversible multiplexer passes through the audio. Each component is then encoded separately.
The bit stream assembler receives the embedded bit stream of the L+R and L-R
channels from the embedded entropy encoders and forms the final bit stream of the
10 system. The bit stream assembler thus allocates the available coding bit rate among
multiple timeslots and channels, truncates the embedded bit stream of each timeslot and channel according to the allocated bit rate, and produces a final compressed bit stream.
To the accomplishment of the foregoing and related ends, certain illustrative
aspects of the invention are described herein in connection with the following description
15 and the annexed drawings. These aspects are indicative, however, of but a few of the
various ways in which the principles of the invention may be employed and the present invention is intended to include all such aspects and their equivalents. Other advantages and novel features of the invention may become apparent from the following detailed description of th§ invention when considered in conjunction with the drawings. 20
BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram of a lossless data coder system in accordance with an aspect of the present invention.
Fig. 2 is a graph illustrating the quantization noise versus rotation angles for
25 different factorization forms in accordance with an aspect of the present invention.
Fig. 3 is a block diagram of a multiple factorization reversible transform component employing an 8-point MLT in accordance with an aspect of the present invention.
Fig. 4 is a block diagram of a lossless data decoder system in accordance with an
30 aspect of the present invention.
7

MS303116.1
Fig. 5 is a block diagram of a progressive to lossless data coder system in accordance with an aspect of the present invention.
Fig. 6 is a block diagram of a progressive to lossless data decoder system in
accordance with an aspect of the present invention.
5 Fig. 7 is a block diagram of a progressive to lossless stereo audio coder system in
accordance with an aspect of the present invention.
Fig. 8 is a block diagram of an exemplary bit stream in accordance with an aspect of the present invention.
Fig. 9 is a block diagram of a progressive to lossless audia decoder system in
10 accordance with an aspect of the present invention.
Fig. 10 is a flow chart flow chart illustrating a method for lossless data encoding in accordance with an aspect of the present invention.
Fig. 11 is a flow chart illustrating a method for lossless data decoding in
accordance with an aspect of the present invention.
15 Fig. 12 is a flow chart illustrating a method for progressive to lossless data
encoding in accordance with an aspect of the present invention.
Fig. 13 is a flow chart illustrating a method for progressive to lossless audio encoding in accordance with an aspect of the present invention.
Fig. 14 is flow chart illustrating a method of multiple factorization reversible
20 transform in accordance with an aspect of the present invention.
Fig. 15 illustrates an example operating environment in which the present invention may function.
DETAILED DESCRIPTION OF THE INVENTION
25 The present invention is now described with reference to the drawings, wherein
like reference numerals are used to refer to like elements throughout. In the following
description, for purposes of explanation, numerous specific details are set forth in order
to provide a thorough understanding of the present invention. It may be evident,
however, that the present invention may be practiced without these specific details. In
30 other instances, well-known structures and devices are shown in block diagram form in
order to facilitate describing the present invention.
8

MS303116.1
As used in this application, the term "computer component" is intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. For example, a computer component may be, but is not limited to being, a process running on a processor, a processor, an object, an
5 executable, a thread of execution, a program, and/or a computer. By way of illustration,
both an application running on a server and the server can be a computer component. One or more computer components may reside within a process and/or thread of execution and a component may be localized on one computer and/or distributed between two or more computers.
10 Referring to Fig. 1, a lossless data coder system 100 in accordance with an aspect
of the present invention is illustrated. The system comprises a multiple factorization reversible transform component 110 and an entropy encoder 120.
The system 100 can losslessly encode data (e.g., image, picture and/or audio). The system 100 receives a digital input (e.g., from an analog to digital converter)
15 associated with, for example, an audio waveform. The system 100 can be employed, for
example, in a vast array of audio and/or document image applications, including, but not limited to, digital audio systems, segmented layered image systems, photocopiers, document scanners, optical character recognition systems, personal digital assistants, fax machines, digital cameras, digital video cameras and/or video games.
20 The multiple factorization reversible transform component 110 receives an input
signal and provides an output of quantized coefficients corresponding to the input signal. The multiple factorization reversible transform component 110 can employ, for example, a reversible modulated lapped transform, a reversible DCT, a reversible DST and/or a reversible FFT.
25 For example, the output of quantized coefficients can be based, at least in part,
upon a reversible modulated lapped transform (MLT). The lossless compression efficiency is affected by the similarity between the reversible MLT and its non-reversible counter part. Therefore, the reversible MLT is designed so that its transform result mimics the result of a non-reversible MLT (e.g., as much as possible). In conventional
30 system(s), the reversible MLT is built with reference to a linear non-reversible MLT, and
by making substantially every component of the non-reversible MLT reversible.
9

MS303116.1
The rotation operation is a key element of a non-reversible MLT, as well as other non-reversible transforms commonly used in compression, such as DCT, DST and/or FFT. To build the reversible MLT, a reversible rotation is built:


(1)

The matrix

cos 0 - sin 0 sin 0 cos 0

is used to denote a reversible rotation operation. A

reversible rotation typically has the following properties:

X from V~
_y_ y.
a. For integer input
of the reversible rotation should be
10

, the outcome quantized coefficients, i.e., integer value as well, b. With proper inverse operation, it is possible to exactly recover

15
20

should bear close resemblance to that of a linear rotation. The
and the linear rotation result is
c. The outcome
difference between the integer outcome defined as the quantization error. Traditionally, a rotation can be factored into a three step non-linear operations:
(2)
where c0=(cos 6-l)/sin 6 and ci=sin 6 are parameters of the operation; [.J denotes an
integerize operation, where the floating operand is converted to an integer number. The rotation clearly is reversible, such that the intermediate integer variable z can be recovered (e.g., exactly) from x' and y' with step 2, and then recover the integer input y from x' and z with step 1, and finally recover integer input x from x and z with step 0.

10

MS303116.1

~xr
y. = 1 c0 0 1 X
y.
Again for notation simplicity, the upper triangle matrix
denote non-linear operation x'= x + [c0yj, and the bottom triangle matrix

is used to


to denote non-linear operation y'= y + [cx]. The reversible rotation
factorization of (2) can thus be denoted as:

10
15


It is to be appreciated that a reversible transform employing the factorization set forth in equation (2) as the basic reversible rotation operation satisfies for the reversible transform yields a reversible transform with compact data representation. However, with certain rotation angle 6, the quantization noise can be fairly large, and can lead to poor signal representation.
Significantly, in accordance with an aspect of the present invention, instead of factoring a linear transform, e.g., a linear rotation into a single reversible factorization form, e.g., in (3), multiple forms of factorization are used. In one example, a proper form of reversible factorization is selected based on one or more transform parameter(s). With multiple factorization, it is possible to achieve much smaller quantization noise, and thus enables the reversible transform to mimic its linear transform counter part as closely as possible. This approach is referred to herein as "multiple factorization reversible transform".
For example, a reversible rotation can be factorized with three additional forms:

20

11

VHS3U3116.1

The core of the factorization is still the three step non-linear operation set forth in equation (3). However, the pair of input/output variables can be swapped before (as in
5 equation (5)) and after (as in equation (4)) the non-linear operation. The sign of the
input/output can be changed as well in certain cases. The additional forms of factorization lead to different non-linear parameters CQ and ci for substantially the same rotation angle 8, and, thus allows the reversible transform to mimic that of a linear non¬reversible transform (e.g., as closely as possible).
10 Let Ax' and Av' represent quantization noise. Quantization noise is the difference
between the outcome of a reversible transform and that of a linear non-reversible transform. In accordance with an aspect of the present invention, a reversible transform that minimizes the average energy of the quantization noise E[Ax"]+E[Ay'z] is desired. The integerize operation introduces the quantization noise into the reversible transform
15 (e.g., based upon truncation and/or rounding). The coefficient swapping and sign
changing operations in equations (4)-(6) do not introduce additional quantization noise. Let A represent the quantization noise of a single integerize operation:


20

(7)

The quantization noise in the reversible transform can be modeled as:


(8)

25 where Ao-A2 are the quantization noise at non-linear steps 0-2. Let the quantization noise
at each step be independent and identically distributed random variables, with £[A2] representing the average energy of the quantization noise of a single integerize operation. The average energy of the quantization noise can be calculated as:
12

MS303116.1
E[Ax'2 J+ E [Ay2 J= {(1 + c0cy +c02 + c\ + 2} E [_A2 J (9)
A graph 200 illustrating the quantization noise versus rotation angles for different
factorization forms as set forth in equations (3) -(6) in accordance with an aspect of the
5 present invention is illustrated in Figure 2. A line 210 denoted by "o" corresponds to
equation (3). A line 220 denoted by "x" corresponds to equation (4). A line 230 denoted by "+" corresponds to equation (5). A line 240 denoted by "0" corresponds to equation (6).
It can be observed that with any single factorization, the quantization noise can be
10 large at certain rotation angle. By switching among different factorizations, more
specifically, by using factorization forms (3), (4), (5) and (6) for rotation angles (-0.25ir,
0.257T), (-0.757T, -0.25*), (0.25ir, 0.75ir) and (0.75x, 1.25x), respectively, the quantization
noise can be controlled to be at most 3.2 £[A2]. In one example, a subset of the
factorization forms (3)-(6) are used, for example, for reduced complexity, albeit at a
15 slightly large quantization noise. For example, factorization form (3) can be used for
rotation angle (-0.5 x, 0.5 TT), and form (6) for rotation angle (0.5 7r,1.5 if). In such a case, the quantization noise is capped at 4 £[A2].
To achieve reduced quantization noise, in one example, rounding towards the
nearest integer can be employed as the integerize operation. In another example,
20 truncation can be employed as the integerize operation. However, rounding toward the
nearest integer can lead to significantly smaller quantization noise and better lossless compression performance, as it has a smaller single step quantization noise £[A2].
A common operation in the reversible operation is a multiplication + integerize
operation in the non-linear operation, which is in the form of [Cy.yJ. In C/C++
25 implementation of reversible transform, the operation is generally implemented as:
floor (ldexp(c/,^*SHIFT_INT) * y + 0.5 );
where y is usually a 32 bit integer, Cf is a floating number represented by a 32 bit integer
30 c, the number Cf is related to the 32 bit integer c through
13

MS303116.1
cf = c*2A(-SHIFT_INT),
where SHIFTINT is a system constant, which is 24 in one example. This can be the
most computational intensive operation of the entire reversible transform. On an x86
system, this can be implemented as inline assembly as:
asm
{

10

MOV eax, y;
IMUL c; // edx:eax = x*y;
SHL edx, SHIFT_REST;
SHR eax, SHIFT INT;
ADC eax, edx;

15
20
25
30

};
where SHIFTREST = 32-SHIFT_rNT. Such operation alone can speed up the implementation, for example, by a factor of 10.
In one example, the multiple factorization reversible transform component 110 can employ a reversible MLT that can be factored into a window modulation and a type IV Discrete Cosine Transform (DCT); the latter of which can be further factored into a pre-Fast Fourier Transform (FFT) rotation, a complex FFT, and a post-FFT rotation operation; each of which can be further built upon multiple factorized reversible rotation.
Referring to Fig. 3, a multiple factorization reversible transform component 300 employing an 8-point MLT in accordance with an aspect of the present invention is illustrated. The MLT 300 comprises a modulation stage 310, a pre-FFT rotation stage 320, a complex FFT stage 330 and a post-FFT rotation stage 340. The multiple factorization reversible transform component 300 is one example of a multiple factorization reversible transform component that can be employed by the multiple factorization reversible transform component 110.
The modulation stage 310 and pre-FFT rotation stage 320 comprise rotation operation as set forth in equation (1) that can be reversibly implemented. The core of the complex FFT stage 330 is the butterfly calculation:


(10)

14

MS303116.1
where xc(i) and yc(i) are complex numbers. The first matrix is a complex rotation that can be reversibly implemented. The second matrix is a complex multiplexer. In this example, a 0.257T rotation for both the real and imaginary part of xc(i) andxc(j) is
implemented. With 0.25* rotation, there is a gain factor of 1/ , thus, the implemented reversible butterfly is:


(ID

In contrast to the butterfly of equation (10) that has an absolute determinant of 2, the
10 absolute determinant of the butterfly of equation (11) is 1, which does not expand the
data and is thus more suitable for lossless compression purpose. In the butterfly
(equation (11)), the output coefficient yc(i) and yc(j) are equally weighted, so that it can be
cascaded in the FFT operation.
The core of the post-FFT rotation stage 340 is the conjugate rotation operation,
15 which can be implemented by changing the sign of the imaginary part after a normal
rotation:

20
25

It is to be appreciated that the transform of equation (12) can be reversibly implemented.
The reversible rotation has a determinant of 1, and the operations of swapping coefficients and changing sign have a determinant of-1. The absolute determinant of the entire reversible MLT is thus 1. Because a determinant 1 transform preserves the signal energy, the quantization noise of the reversible MLT is roughly proportional to the number of rotations in the transform, which is 0(Nlog2N)E[A2] for an N-point MLT (note that there are two rotations per butterfly in the complex FFT). Such implementation is more favorable than factoring the reversible MLT (or type IV DCT) through an N-point LU transform, where the quantization noise that is caused by a long chain of non-linear operation can be much larger.

15

MS303116.1
Referring back to Fig. 1, the entropy encoder 120 digitally entropy encodes the quantized coefficients. The entropy encoder 120 can employ any suitable entropy encoding technique(s). Accordingly, any suitable entropy encoding technique for employment in connection with the present invention is intended to fall within the scope
5 of the appended claims.
While Fig. 1 is a block diagram illustrating components for the lossless data coder system 100 that uses a multiple factorization reversible transform components. It is to be appreciated that the lossless data coder system 100, the multiple factorization reversible transform component 110 and/or the entropy encoder 120 can be implemented as one or
10 more computer components, as that term is defined herein. Thus, it is to be appreciated
that computer executable components operable to implement the lossless data coder system 100, the multiple factorization reversible transform component 110 and/or the entropy encoder 120 can be stored on computer readable media including, but not limited to, an ASIC (application specific integrated circuit), CD (compact disc), DVD (digital
15 video disk), ROM (read only memory), floppy disk, hard disk, EEPROM (electrically
erasable programmable read only memory) and memory stick in accordance with the present invention.
Next, referring to Fig. 4, a lossless data decoder system 400 in accordance with an aspect of the present invention is illustrated. The system 400 comprises an entropy
20 decoder 410 and an inverse transform component 420.
The entropy decoder 410 receives a bit stream (e.g., produced by a corresponding entropy encoder) and decodes the bit stream. Thus, the entropy decoder 410 digitally entropy decodes the input bit stream and provides the decoded information to the inverse transform component 420.
25 The inverse transform component 420 receives output values from the entropy
decoder 410. The inverse transform component 420 transforms the output values from the entropy decoder 410 and provides output values. In one example, the inverse transform component 420 utilizes an inverse reversible MLT, to essentially revert the computations in the multiple factorization reversible transform component 110.
16

MS303116.1
It is to be appreciated that the lossless data decoder system 400, the entropy decoder 410 and/or the inverse transform component 420 can be computer components as that term is defined herein.
Turning to Fig. 5, a progressive to lossless data coder system 500 in accordance
5 with an aspect of the present invention is illustrated. The system 500 comprises a
multiple factorization reversible transform component 110 and an embedded entropy encoder 510.
The progressive to lossless data coder system 500 can be employed to code audio and/or images. In one example, the system 500 is based on reversible transform that is
10 designed to mimic the non-reversible transform in a normal psychoacoustic audio coder
(e.g., as much as possible). The multiple factorization reversible transform component 110 coupled with the embedded entropy encoder 510 (e.g., a high performance embedded entropy codec) empowers the system 500 with both lossless capability and fine granular scalability progressiveness. The system 500 generates a bit stream that if fully decoded,
15 substantially recovers the original audio waveform without loss and facilitates lossless
compression. If a higher compression ratio is desired, the application may extract a subset of the compressed bit stream and form a higher compression ratio bit stream of lossy nature. Such scaling can be performed in a very large bit rate range, with granularity down to a single byte. With the progressive to lossless functionality of the
20 system 500, the application can easily balance between the amount of compression
required and the desired audio quality, from a fairly high compression ratio all the way to lossless (e.g., based on available bandwidth and/or user preference).
The embedded entropy encoder 510 can employ any suitable embedded entropy encoding technique(s). Accordingly, any suitable embedded entropy encoding technique
25 for employment in connection with the present invention is intended to fall within the
scope of the appended claims.
In one example, the embedded entropy encoder 510 employs a fully scalable psychoacoustic audio coder that derives auditory masking thresholds from previously coded coefficients, and uses the derived thresholds for optimizing the order of coding.
30 The basic theory behind acoustic or auditory masking is described in general terms
herein.
17

MS303116.1
In general, the basic theory behind auditory masking is that humans do not have the ability to hear minute differences in frequency. For example, it is very difficult to discern the difference between a 1,000 Hz signal and a signal that is 1,001 Hz. It becomes even more difficult for a human to differentiate such signals if the two signals
5 are playing at the same time. Further, studies have shown the 1,000 Hz signal would also
affect a human's ability to hear a signal that is 1,010 Hz, or 1,100 Hz, or 990 Hz. This concept is known as masking. If the 1,000 Hz signal is strong, it will mask signals at nearby frequencies, making them inaudible to the listener. In addition, there are two other types of acoustic masking which affects human auditory perception. In particular,
10 as discussed below, both temporal masking and noise masking also effect human audio
perception. These ideas are used to improve audio compression because any frequency components in the audio file which fall below a masking threshold can be discarded, as they will not be perceived by a human listener.
In this example, the embedded entropy encoder 510 is a fully scalable generic
15 audio coder that uses implicit auditory masking that is intermixed with a scalable entropy
coding process. Further, auditory masking thresholds are not sent to the decoder, instead, they are derived from the already coded coefficients. Furthermore, rather than quantizing the audio coefficients according to the auditory masking thresholds, the masking thresholds are used to control the order that the coefficients are encoded. In particular,
20 during the scalable coding, larger audio coefficients are encoded first, as the larger
components are the coefficients that contribute most to the audio energy level and lead to a higher auditory masking threshold.
For example, the MLT transform coefficients received from the multiple factorization reversible transform component 110 are split into a number of sections.
25 This section split operation enables the scalability of the audio sampling rate. Such
scalability is particularly useful where different frequency responses of the decoded audio file are desired. For example, where one or more playback speakers associated with the decoder do not have a high frequency response, or where it is necessary for the decoder to save either or both computation power and time, one or more sections corresponding to
30 particular high frequency components of the MLT transform coefficients can be
discarded.
18

MS303116.1
Each section of the MLT transform coefficients is then entropy encoded into an
embedded bit stream, which can be truncated and reassembled at a later stage. Further, to
improve the efficiency of the entropy coder, the MLT coefficients are grouped into a
number of consecutive windows termed a timeslot. In a one example, a timeslot consists
5 of 16 long MLT windows or 128 short MLT windows. However, it should be clear to
those skilled in the art that the number of windows can easily be changed.
In particular, first, the auditory masking is used to determine the order that the transform coefficients are encoded, rather than to change the transform coefficients by quantizing them. Instead of coding any auditory insensitive coefficients coarsely, in this
10 example, the embedded entropy coder 510 encodes such coefficients in a later stage. By
using the auditory masking to govern the coding order, rather than the coding content, the embedded entropy coder 510 achieves embedded coding up to and including lossless encoding of the audio input, as all content is eventually encoded. Further, the quality of the audio becomes less sensitive to the auditory masking, as slight inaccuracies in the
15 auditory masking simply cause certain audio coefficients to be encoded later.
Second, in this example, the auditory masking threshold is derived from the already encoded coefficients, and gradually refined with the embedded coder (e.g., "implicit auditory masking"). In implementing the implicit audio masking, the most important portion of the transform coefficients (e.g., the top bitplanes), are encoded first.
20 A preliminary auditory masking threshold is calculated based on the already coded
transform coefficients. Since the decoder automatically derives the same auditory masking threshold from the coded transform coefficients, the value of the auditory masking threshold does not need to be sent to the decoder. Further, the calculated auditory masking threshold is used to govern which part of the transform coefficients is
25 to be refined.
After the next part of the transform coefficients has been encoded, a new set of auditory masking threshold is calculated. This process repeats until a desired end criterion has been met, e.g., all transform coefficients have been encoded, a desired coding bit rate has been reached, or a desired coding quality has been reached. By
30 deriving the auditory masking threshold from the already coded coefficients, bits
19

MS303116.1
normally required to encode the auditory masking threshold are saved. Consequently, the coding quality is improved, especially when the coding bit rate is low.
In another example, the embedded entropy encoder 510 employs a traditional
coder which carries the auditory masking threshold as a head of the bit stream.
5 Thus, after the reversible transform is performed by the multiple factorization
reversible transfonn component 110, the MLT coefficients of multiple windows are grouped into a timeslot. The coefficients of each timeslot are then entropy encoded by the embedded entropy encoder 510 which not only efficiently compresses the coefficients, but also renders die output bit stream with the embedding (e.g., progressive)
10 property, so that the bit stream (e.g., of each channel) can be truncated at any point.
It is to be appreciated that the system 500 and/or the embedded entropy encoder 510 can be computer components as that term is defined herein.
Referring to Fig. 6, a progressive to lossless data decoder system 600 in accordance with an aspect of the present invention is illustrated. The system comprises
15 an embedded entropy decoder 610 and an inverse transform component 420.
The embedded entropy decoder 610 receives a bit stream (e.g., produced by a corresponding entropy encoder) and decodes the bit stream. Thus, the embedded entropy decoder 610 digitally entropy decodes the input bit stream and provides the decoded information to the inverse transfonn component 420.
20 The inverse transform component 420 receives output values from the embedded
entropy decoder 610. The inverse transform component 420 transforms the output values from the embedded entropy decoder 610 and provides output values. In one example, the inverse transform component 420 utilizes an inverse reversible MLT, to essentially revert the computations in the multiple factorization reversible transform component 110.
25 It is to be appreciated that the progressive to lossless data decoder system 600
and/or the embedded entropy decoder 610 can be computer components as that term is defined herein.
Next, turning to Fig. 7, a progressive to lossless stereo audio coder system 700 in accordance with an aspect of the present invention is illustrated. The system 700
30 comprises a reversible multiplexer 710, multiple factorization reversible transform
20

MS303116.1

components 110i, 1102, embedded entropy encoders 510i, 5102 and a bit stream assembler 720.
The system 700 receives an input audio waveform which first goes through the
reversible multiplexer 710. If the input audio is stereo, it is separated into L + R and L -
5 R components. The waveform of each audio component is then transformed by the
multiple factorization reversible transform components 110j, 1IO2 (e.g., with switching windows). In one example, the window size can be either 2048 or 256 samples. After the multiple factorization reversible transform components 110i, 1IO2, the MLT coefficients of a number of consecutive windows are grouped into a timeslot.
10 For example, a timeslot can consist of 16 long MLT windows or 128 short
windows. A timeslot therefore consists of 32,768 samples, which is about 0.74 second if the input audio is sampled at 44.1 kHz. The coefficients in the timeslot are then entropy encoded by the embedded entropy encoders 510i, 5IO2 whose output bit stream can be truncated at any point later.
15 The reversible multiplexer 710 receives, for example, a stereo audio input. The
reversible multiplexer 710 separates the stereo audio input into L+R and L-R components using conventional techniques, where L and R represent the waveform on the left and right audio channel, respectively. If the input audio is mono, the reversible multiplexer passes through the audio. Each component is then encoded separately.
20 Let x and y be the left and right channel, x' and y' be the multiplexed channel L+R
and L-R, a reversible multiplexer can be implemented in non-linear form as:

(13)
25 or alternatively, be written as

where \x\ again denotes an integerize operation. Equation (13) produces integer input
30 from integer output and can be exactly reversed. Ignoring the nonlinearity in the
21

MS303116.1
integerize operation, the relationship between the input/output pair can be formulated through a linear transform as:


(15)

The determinant of the transform of equation (15) is -1, which means the volume of the input/output space is equal.
There are alternative implementations of the reversible multiplexer, e.g., with a rotation by 0.25 x operation, or via:
10 (16)
However, the reversible multiplexer in the form of equation (16) has an absolute determinant of 2, which expands the data and is less efficient for lossless compression. The reversible multiplexer using 0.25 7r rotation operation needs 3 integerize operations and has a quantization noise of 3.2 E[A2], the reversible multiplexer of equation (13) has
15 only one integerize operation, with a quantization noise of£[Az]. It is thus a more
favorable implementation. With the reversible multiplexer of equation (13), the L-R and L+R channels are not equally weighted, the L-R channels carries a weight which is 4 times heavier than that of the L+R channels. The non-equal weight is not problematic in lossless coding, as all quantized coefficients of the L-R and L+R channels are simply
20 encoded. In case the lossless bit stream is truncated, it can compensate the non-equal
weighting by encoding the L+R channels with one more bitplane.
The bit stream assembler 720 receives the embedded bit stream of the L+R and L-R channels from the embedded entropy encoders 510i, 5102. The bit stream assembler forms the final bit stream of the system 700.
25 An exemplary bit stream 800 in accordance with an aspect of the present
invention is illustrated in Fig. 8. The bit stream 800 comprises a global header 810, which is followed by a first timeslot 820i through an N* timeslot 820N, N being an integer greater than or equal to one. Each timeslot 820 comprises a timeslot header 830, which records the length of the compressed bit stream in the L+R and L-R channels, and
22

MS303116.1
is followed by the actual embedded bit stream of the channels 840, 850. If exact
waveform reconstruction is required, the entire bit stream will be decoded. In case higher
compression ratio is called for, a subset from the losslessly encoded bit stream to form a
bit stream of higher compression ratio is extracted. Since this is achieved by truncating
5 the embedded bit stream of the L+R and L-R channels of individual timeslot, the
operation can be performed very fast. It is also possible to convert the compressed audio from stereo to mono by removing the compressed bit stream associated with the L-R channel.
Turning back to Fig. 7, the bit stream assembler 720 thus allocates the available
10 coding bit rate among multiple timeslots and channels, truncates the embedded bit stream
of each timeslot and channel according to the allocated bit rate, and produces a final compressed bit stream.
It is to be appreciated that the system 700, the reversible multiplexer 710 and/or the bit stream assembler 720 can be computer components as that term is defined herein.
15 Referring next to Fig. 9, a progressive to lossless audio decoder system 900 in
accordance with an aspect of the present invention is illustrated. The system 900 comprises a bit stream unassembler 910, embedded entropy decoders 610j, 6IO2, inverse transform components 420i, 4202 and a reversible demultiplexer 920.
The bit stream unassembler 910 receives an input bit stream (e.g., from a
20 corresponding bit stream assembler 720). The bit stream unassembler separates the L+R
channel and the L-R channel and provides the separated channels to the embedded entropy decoders 610j, 6102.
Each of the embedded entropy decoders 6IO1,6IO2 receives a channel of information from the bit stream unassembler 910 and decodes the bit stream. Thus, the
25 embedded entropy decoders 6101,6102 digitally entropy decodes the input bit stream and
provides the decoded information to the inverse transform components 4201,4202.
The inverse transform components 4201,4202 receive output values from the embedded entropy decoders 6IO1,6IO2. The inverse transform components 420i, 4202 transform the output values from the embedded entropy decoders 6IO1,6IO2 and provides
30 output values. In one example, the inverse transform components 420], 4202 utilize an
23

MS303116.1
inverse reversible MLT, to essentially revert the computations in the multiple
factorization reversible transform component 110.
The reversible demultiplexer 920 receives the outputs of the inverse transform
components 420i, 4202 and demultiplexes the outputs into a stereo audio waveform.
5 It is to be appreciated that the system 900, the bit stream unassembler 910 and/or
the reversible demultiplexer 920 can be computer components as that term is defined herein.
Turning briefly to Figs. 10,11, 12,13 and 14, methodologies that may be implemented in accordance with the present invention are illustrated. While, for
10 purposes of simplicity of explanation, the methodologies are shown and described as a
series of blocks, it is to be understood and appreciated that the present invention is not limited by the order of the blocks, as some blocks may, in accordance with the present invention, occur in different orders and/or concurrently with other blocks from that shown and described herein. Moreover, not all illustrated blocks may be required to
15 implement the methodologies in accordance with the present invention.
The invention may be described in the general context of computer-executable instructions, such as program modules, executed by one or more components. Generally, program modules include routines, programs, objects, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically the functionality of
20 the program modules may be combined or distributed as desired in various embodiments.
Referring to Fig. 10, a method for lossless data encoding 1000 in accordance with an aspect of the present invention is illustrated. At 1010, an input signal is received {e.g., audio and/or image). At 1020, quantized coefficients corresponding to the input signal based, at least in part, upon a multiple factorization reversible transform (e.g., reversible
25 modulated lapped transform) is provided. For example, the reversible modulated lapped
transform can be based upon equations (3), (4), (5), (6), (9), (10) and/or (11). The multiple factorization reversible transform can comprise a modulation stage, a pre-FFT rotation stage, a complex FFT stage and/or a post-FFT rotation stage. At 1030, the quantized coefficients are entropy encoded.
30 Next, turning to Fig. 11, a method for lossless data decoding 1100 in accordance
with an aspect of the present invention is illustrated. At 1110, a digitally entropy
24

MS303116.1
encoded input bit stream is received. At 1120, the bit stream is entropy decoded and
transform coefficients are provided. At 1130, output values based on an inverse
transform of the transform coefficients are provided. The inverse transform can be based
on a reversible modulated lapped transform.
5 Referring to Fig. 12, a method for progressive to lossless data encoding 1200 in
accordance with an aspect of the present invention is illustrated. At 1210, an input signal is received (e.g., audio and/or image). At 1220, quantized coefficients corresponding to the input signal based, at least in part, upon a multiple factorization reversible transform (e.g., reversible modulated lapped transform) is provided. For example, the multiple
10 factorization reversible transform can be based upon equations (3), (4), (5), (6), (9), (10)
and/or (11). The multiple factorization reversible transform can comprise a modulation stage, a pre-FFT rotation stage, a complex FFT stage and/or a post-FFT rotation stage. At 1230, a progressive bit stream based upon embedded encoding of the quantized coefficients is provided.
15 Turning to Fig. 13, a method for progressive to lossless audio encoding 1300 in
accordance with an aspect of the present invention is illustrated. At 1310, an input signal is received. At 1320, the input signal is separated into a first component (L+R) and a second component (L-R) (e.g., by a reversible multiplexer 710).
At 1330, quantized coefficients corresponding to the first component based, at
20 least in part, upon a multiple factorization reversible transform (e.g., reversible
modulated lapped transform) are provided. For example, the multiple factorization reversible lapped transform can be based upon equations (3), (4), (5), (6), (9), (10) and/or (11). The multiple factorization reversible transform can comprise a modulation stage, a pre-FFT rotation stage, a complex FFT stage and/or a post-FFT rotation stage. At 1340, a
25 progressive bit stream based upon embedded encoding of the quantized coefficients
associated with the first component is provided.
At 1350, quantized coefficients corresponding to the second component based, at least in part, upon a multiple factorization reversible transform (e.g., reversible modulated lapped transform) are provided. For example, the multiple factorization
30 reversible transform can be based upon equations (3), (4), (5), (6), (9), (10) and/or (11).
The multiple factorization reversible transform can comprise a modulation stage, a pre-
25

MS303U6.1
FFT rotation stage, a complex FFT stage and/or a post-FFT rotation stage. At 1360, a progressive bit stream based upon embedded encoding of the quantized coefficients associated with the second component is provided.
At 1370, the progressive bit stream of the embedded entropy encoded quantized
5 coefficients associated with the first and second component are assembled into an output
bit stream (e.g., by a bit stream assembler 720).
Referring to Fig. 14, a method of multiple factorization reversible transform 1400 in accordance with an aspect of the present invention is illustrated. At 1410, at least one factorization of the multiple factorization reversible transform is performed. At 1420, a
10 factorization which produces a minimum amount of quantization noise based, at least in
part, upon a parameter of the transform, is selected. For example, the factorization can be based upon equations (3), (4), (5), (6), (9), (10) and/or (11). The factorization can comprise a modulation stage, a pre-FFT rotation stage, a complex FFT stage and/or a post-FFT rotation stage as set forth supra.
15 In order to provide additional context for various aspects of the present invention,
Fig. 15 and the following discussion are intended to provide a brief, general description of a suitable operating environment 1510 in which various aspects of the present invention may be implemented. While the invention is described in the general context of computer-executable instructions, such as program modules, executed by one or more
20 computers or other devices, those skilled in the art will recognize that the invention can
also be implemented in combination with other program modules and/or as a combination of hardware and software. Generally, however, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular data types. The operating environment 1510 is only one example of
25 a suitable operating environment and is not intended to suggest any limitation as to the
scope of use or functionality of the invention. Other well known computer systems, environments, and/or configurations that may be suitable for use with the invention include but are not limited to, personal computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, programmable consumer
30 electronics, network PCs, minicomputers, mainframe computers, distributed computing
environments that include the above systems or devices, and the like.
26

MS303116.1
With reference to Fig. 15, an exemplary environment 1510 for implementing various aspects of the invention includes a computer 1512. The computer 1512 includes a processing unit 1514, a system memory 1516, and a system bus 1518. The system bus 1518 couples system components including, but not limited to, the system memory 1516
5 to the processing unit 1514. The processing unit 1514 can be any of various available
processors. Dual microprocessors and other multiprocessor architectures also can be employed as the processing unit 1514.
The system bus 1518 can be any of several types of bus structure^) including the memory bus or memory controller, a peripheral bus or external bus, and/or a local bus
10 using any variety of available bus architectures including, but not limited to, 16-bit bus,
Industrial Standard Architecture (ISA), Micro-Channel Architecture (MSA), Extended ISA (EISA), Intelligent Drive Electronics (IDE), VESA Local Bus (VLB), Peripheral Component Interconnect (PCI), Universal Serial Bus (USB), Advanced Graphics Port (AGP), Personal Computer Memory Card International Association bus (PCMCIA), and
15 Small Computer Systems Interface (SCSI).
The system memory 1516 includes volatile memory 1520 and nonvolatile memory 1522. The basic input/output system (BIOS), containing the basic routines to transfer information between elements within the computer 1512, such as during start-up, is stored in nonvolatile memory 1522. By way of illustration, and not limitation,
20 nonvolatile memory 1522 can include read only memory (ROM), programmable ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory 1520 includes random access memory (RAM), which acts as external cache memory. By way of illustration and not limitation, RAM is available in many forms such as synchronous RAM (SRAM), dynamic RAM
25 (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM),
enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM(DRRAM).
Computer 1512 also includes removable/nonremovable, volatile/nonvolatile computer storage media. Fig. 15 illustrates, for example a disk storage 1524. Disk
30 storage 1524 includes, but is not limited to, devices like a magnetic disk drive, floppy
disk drive, tape drive, Jaz drive, Zip drive, LS-100 drive, flash memory card, or memory
27

MS303116.1
stick. In addition, disk storage 1524 can include storage media separately or in
combination with other storage media including, but not limited to, an optical disk drive
such as a compact disk ROM device (CD-ROM), CD recordable drive (CD-R Drive), CD
rewritable drive (CD-RW Drive) or a digital versatile disk ROM drive (DVD-ROM). To
5 facilitate connection of the disk storage devices 1524 to the system bus 1518, a
removable or non-removable interface is typically used such as interface 1526.
It is to be appreciated that Fig 15 describes software that acts as an intermediary between users and the basic computer resources described in suitable operating environment 1510. Such software includes an operating system 1528. Operating system
10 1528, which can be stored on disk storage 1524, acts to control and allocate resources of
the computer system 1512. System applications 1530 take advantage of the management of resources by operating system 1528 through program modules 1532 and program data 1534 stored either in system memory 1516 or on disk storage 1524. It is to be appreciated that the present invention can be implemented with various operating systems
15 or combinations of operating systems.
A user enters commands or information into the computer 1512 through input device(s) 1536. Input devices 1536 include, but are not limited to, a pointing device such as a mouse, trackball, stylus, touch pad, keyboard, microphone, joystick, game pad, satellite dish, scanner, TV tuner card, digital camera, digital video camera, web camera,
20 and the like. These and other input devices connect to the processing unit 1514 through
the system bus 1518 via interface port(s) 1538. Interface port(s) 1538 include, for example, a serial port, a parallel port, a game port, and a universal serial bus (USB). Output device(s) 1540 use some of the same type of ports as input device(s) 1536. Thus, for example, a USB port may be used to provide input to computer 1512, and to output
25 information from computer 1512 to an output device 1540. Output adapter 1542 is
provided to illustrate that there are some output devices 1540 like monitors, speakers, and printers among other output devices 1540 that require special adapters. The output adapters 1542 include, by way of illustration and not limitation, video and sound cards that provide a means of connection between the output device 1540 and the system bus
30 1518. It should be noted that other devices and/or systems of devices provide both input
and output capabilities such as remote computers) 1544.
28

MS303116.1
Computer 1512 can operate in a networked environment using logical connections
to one or more remote computers, such as remote computers) 1544. The remote
computers) 1544 can be a personal computer, a server, a router, a network PC, a
workstation, a microprocessor based appliance, a peer device or other common network
5 node and the like, and typically includes many or all of the elements described relative to
computer 1512. For purposes of brevity, only a memory storage device 1546 is illustrated with remote computer(s) 1544. Remote computers) 1544 is logically connected to computer 1512 through a network interface 1548 and then physically connected via communication connection 1550. Network interface 1548 encompasses
10 communication networks such as local-area networks (LAN) and wide-area networks
(WAN). LAN technologies include Fiber Distributed Data Interface (FDDI), Copper Distributed Data Interface (CDDI), Ethernet/IEEE 802.3, Token Ring/IEEE 802.5 and the like. WAN technologies include, but are not limited to, point-to-point links, circuit switching networks like Integrated Services Digital Networks (ISDN) and variations
15 thereon, packet switching networks, and Digital Subscriber Lines (DSL).
Communication connection(s) 1550 refers to the hardware/software employed to connect the network interface 1548 to the bus 1518. While communication connection 1550 is shown for illustrative clarity inside computer 1512, it can also be external to computer 1512. The hardware/software necessary for connection to the network interface
20 1548 includes, for exemplary purposes only, internal and external technologies such as,
modems including regular telephone grade modems, cable modems and DSL modems, ISDN adapters, and Ethernet cards.
What has been described above includes examples of the present invention. It is, of course, not possible to describe every conceivable combination of components or
25 methodologies for purposes of describing the present invention, but one of ordinary skill
in the art may recognize that many further combinations and permutations of the present invention are possible. Accordingly, the present invention is intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims. Furthermore, to the extent that the term "includes" is used in either the
30 detailed description or the claims, such term is intended to be inclusive in a manner
29

MS303116.1
CLAIMS
What is claimed is:
1. A lossless data coder system comprising:
a multiple factorization reversible transform component that receives an input signal and provides an output of quantized coefficients corresponding to the input signal, the output of quantized coefficients being based, at least in part, upon a multiple factorization reversible transform; and,
an entropy encoder that digitally entropy encodes the quantized coefficients.
2. The system of claim 1, wherein the multiple factorization reversible transform is a multiple factorization reversible MLT.
3. The system of claim 2, wherein the multiple factorization reversible MLT further comprises a modulation stage, a pre-FFT rotation stage, a complex FFT stage and a post-FFT rotation stage.
4. The system of claim 3, wherein the modulation stage is implemented via a multiple factorization reversible rotation:

where #is an angle of rotation.
5. The system of claim 3, wherein the pre-FFT rotation is implemented via multiple factorization reversible rotation:

where 0is an angle of rotation.
31


MS303116.1
The system of claim 1, wherein the complex FFT is implemented via:
where xc(i) is a complex number representation of a first channel of the input signal,
jtc(j) is a complex number representation of a second channel of the input signal,