US6636830B1 — System and method for noise reduction using bi-or…

Claim 1

Independent Claim 78 words

A method for encoding a signal, comprising the steps of: transforming the signal into a plurality of BMDCT frequency coefficients using a bi-orthogonal modified discrete cosine transform, wherein three different BMDCT transform sizes are being used; quantizing the BMDCT frequency coefficients to produce a set of integer numbers which represent the BMDCT frequency coefficients; and losslessly encoding the set of integer numbers into representative symbols to lower the number of bits required to represent the BMDCT frequency coefficients.

Claim 2

Independent Claim 78 words

A method for encoding a signal, comprising the steps of: transforming the signal into a plurality of BMDCT frequency coefficients using a bi-orthogonal modified discrete cosine transform, wherein three different BMDCT transform sizes are being used; quantizing the BMDCT frequency coefficients to produce a set of integer numbers which represent the BMDCT frequency coefficients; and losslessly encoding the set of integer numbers into representative symbols to lower the number of bits required to represent the BMDCT frequency coefficients.

Claim 3

Dependent Claim 18 words

The method of claim 1 , wherein the three different BMDCT transform sizes are 512, 256 and 128.

Claim 4

Dependent Claim 29 words

The method of claim 1 , wherein signal analysis is performed on current and future audio frames to determine the optimal window size to use for the current frame.

Claim 5

Dependent Claim 25 words

An encoding method of claim 1 , wherein the BMDCT window coefficients at the decoder are represented at a reduced number of bits of precision.

Claim 6

Dependent Claim 89 words

The method of claim 4 , wherein the decoder BMDCT window coefficients are given by w s = w s 2 nbits where w s   ( n ) = round   ( ( 2 nbits - 1 )   sin  [ π 2  M   ( n + 0.5 ) ] ) , n = 0   …   2  M - 1 , where M is the transform size and nbits is the number of bits of precision.

Claim 7

Dependent Claim 31 words

An encoding method of claim 1 , wherein the BMDCT window coefficients at the encoder are designed such that the combination of forward transform and inverse transform perfectly reconstruct the signal.

Claim 8

Dependent Claim 79 words

The method of claim 6 , wherein the encoder BMDCT window is defined as w a ( n )= w s ′ ( n )/ c ( n ), n =0 . . . M −1 w a (2 M −1 −n )= w a ( n ), n =0 . . . M −1 where c ( n )= w s ′2 ( n )+ w s ′2 ( M+n ), n =0 . . . M −1.

Claim 9

Dependent Claim 33 words

The method of claim 1 , wherein the BMDCT coefficients are quantized using uniform quantization such that each quantized coefficient is represented as C q =Round( C /step_size), for a given step size.

Claim 10

Dependent Claim 17 words

The method of claim 8 , wherein the quantization step sizes vary with frequency and with time.

Claim 11

Dependent Claim 61 words

The method of claim 8 , wherein the quantization step further includes frequency variation of the step sizes by dividing the frequency spectrum into a set of bands over which the quantization step size is constant and grouping one or more sets of BMDCT transform coefficients together into a frame that uses the same step sizes as a function of frequency.

Claim 12

Dependent Claim 56 words

The method of claim 8 , wherein quantization step sizes are calculated by determining a set of transparent noise levels using a psycho-acoustic model, representing noise that can be introduced into the signal and just remain inaudible, and calculating quantization step sizes such that the introduced quantization noise is at or below the transparent noise levels.

Claim 13

Dependent Claim 122 words

The method of claim 11 , wherein the psycho-acoustic model uses tonality measure that is calculated as t = 1.0 - [ - ∑   e   ( i )   log   10   ( e   ( i ) ) ∑   e   ( i ) + log   10   ( ∑   e   ( i ) ) log   10   ( w ) ] where w is the width of a band of coefficients used to calculate tonality and e(i) are the spectral line energies, and wherein the sums are calculated over all the BMDCT coefficients that fall inside the band.

Claim 14

Dependent Claim 149 words

The method of claim 11 , wherein the quantization step sizes are calculated from the desired noise levels, wherein the calculation comprises the steps of: setting the initial step size to the closest value to {square root over (12nT/width)}, wherein nT is the noise target and width is the width of the frequency band over which the desired noise level has been calculated; using this step size to calculate the actual noise which is introduced in the band by quantizing the signal and taking the sum squared difference between the quantized signal and the original signal; reducing the step size in small decrements until the actual noise is less than the target if the noise introduced is greater than the target noise; and increasing the step size in small increments until the noise is just below or equal to the target if the noise is less than the target.

Claim 15

Dependent Claim 46 words

The method of claim 11 , wherein the quantization step sizes for a frame are adjusted by adjusting the target noise level for the frame, whereby the total number of bits required to code the frame is equal to or just less than a given target.

Claim 16

Dependent Claim 33 words

The method of claim 14 , wherein the target number of bits for a frame is determined from the complexity of the current frame in relation to the average complexity of previous frames.

Claim 17

Dependent Claim 71 words

The method of claim 15 , wherein the target number of bits for a frame, n, is given by tBits   ( n ) = avBits · { w + ( 1 - w )   frameComplexity   ( n ) avComplexity   ( n ) } where w is a number greater than 0.5 and less than 1.0 and avBits is a constant bit rate target.

Claim 18

Dependent Claim 56 words

The method of claim 16 , wherein the frame complexity is given by the total number of bits required to code the frame using the transparent noise target, wherein average complexity is calculated using the recursive equation avComplexity( n )=(1−ν)·avComplexity( n −1)+ν·frameComplexity( n ) where ν is a constant greater than 0 and less than 0.5.

Claim 19

Dependent Claim 119 words

The method of claim 14 , wherein the target noise level for a frame is adjusted by linearly interpolating between the transparent noise level and the target noise curve tNoise  [ b ] = K   ( noise  [ b ] energy  [ b ] ) rcWeight   energy  [ b ] where K is a constant, noise[b] are the transparent noise levels for step band b, energy[b] are the signal energies in band b and rcWeight is a constant which determines the shape of the target noise curve in the case where the target number of bits is lower than the number of bits produced by coding using the transparent noise levels.

Claim 20

Dependent Claim 63 words

The method of claim 1 , wherein the quantized BMDCT coefficients are organized into a frame consisting of two (2) blocks of up to 512 coefficients, or four (4) blocks of up to 256 coefficients, or eight (8) blocks of up to 128 coefficients, wherein the coefficients from different blocks are grouped and scanned flexibly into a single array of quantized BMDCT coefficients.

Claim 21

Dependent Claim 29 words

The method of claim 1 , wherein losslessly encoding the quantized BMDCT coefficients includes losslessly encoding the integers into representative symbols using fixed and variable length lossless encoding methods.

Claim 22

Dependent Claim 40 words

The method of claim 20 , wherein each symbol created by the lossless encoding method can represent one or more coefficients in a given range of values and wherein for each method a unique end of block symbol is defined.

Claim 23

Dependent Claim 64 words

The method of claim 21 , wherein a final bit stream is created that consists of repeated occurrences of index symbols, coefficients symbols and end of block symbols, until all coefficients are encoded, wherein the index symbol represents the particular lossless encoding method used and defines the mapping of coefficient symbols occurring between the index and the end of block symbol into coefficient values.

Claim 24

Dependent Claim 42 words

The method of claim 22 , wherein the coefficient symbols are created using fixed length encoding methods derived from Pulse Code Modulation (PCM) encoders, and wherein the maximum positive PCM symbol value generated is used to represent the end of block symbol.

Claim 25

Dependent Claim 18 words

The method of claim 22 , wherein the variable length lossless encoding methods are performed by Huffman encoders.

Claim 26

Independent Claim 170 words

A method for encoding a stereo input signal having a plurality of channels, comprising the steps of: transforming a signal from each of the plurality of channels of the stereo signal into a plurality of BMDCT frequency coefficients using a bi-orthogonal modified discrete cosine transform, wherein transforming a signal from each of the plurality of channels includes performing 2×2 orthogonal transforms on two channels of the stereo signal into independent sets of BMDCT coefficients: [ L ′ R ′ ] = [ cos   ( θ ) sin   ( θ ) - sin   ( θ ) cos   ( θ ) ] · [ L R ] ; quantizing the BMDCT frequency coefficients to produce a set of integer numbers which represent the BMDCT frequency coefficients generated from each channel; and losslessly encoding the set of integer numbers into representative symbols to lower the number of bits required to represent the BMDCT frequency coefficients generated from the plurality of channels of the stereo signal.

Claim 27

Independent Claim 170 words

A method for encoding a stereo input signal having a plurality of channels, comprising the steps of: transforming a signal from each of the plurality of channels of the stereo signal into a plurality of BMDCT frequency coefficients using a bi-orthogonal modified discrete cosine transform, wherein transforming a signal from each of the plurality of channels includes performing 2×2 orthogonal transforms on two channels of the stereo signal into independent sets of BMDCT coefficients: [ L ′ R ′ ] = [ cos   ( θ ) sin   ( θ ) - sin   ( θ ) cos   ( θ ) ] · [ L R ] ; quantizing the BMDCT frequency coefficients to produce a set of integer numbers which represent the BMDCT frequency coefficients generated from each channel; and losslessly encoding the set of integer numbers into representative symbols to lower the number of bits required to represent the BMDCT frequency coefficients generated from the plurality of channels of the stereo signal.

Claim 28

Dependent Claim 30 words

The method of claim 25 , wherein the 2×2 orthogonal transform varies as a function of time and frequency, and are represented as transform angles substantially between 0 and π/2.

Claim 29

Dependent Claim 42 words

The method of claim 25 , wherein redundancy is removed from the two channels of the stereo signal by performing 2×2 orthogonal transforms, wherein the transform angle is calculated by taking the eigenvectors of the cross-correlation matrix calculated from the BMDCT outputs.

Claim 30

Dependent Claim 85 words

The method of claim 25 , wherein irrelevant information is removed from the two channels of the stereo signal by summing or differencing the two channels above a certain minimum frequency to form one channel and setting the other channel to zero above the minimum frequency, and includes substantially simultaneously calculating a set of angles as θ=tan −1 ( E r /E 1 ) where E r is the right channel (i.e. second channel) energy, and E l is the left channel (first channel) energy.

Claim 31

Dependent Claim 76 words

The method of claim 25 , wherein the two channels of the stereo signal are reconstituted at the decoder by performing inverse 2×2 orthogonal transforms [ L R ] = [ cos   ( θ ) - sin   ( θ ) sin   ( θ ) cos   ( θ ) ] · [ L ′ R ′ ] where θ represents transform angles received as side information from the encoder.

Claim 32

Independent Claim 233 words

An apparatus configured to encode a signal, comprising: a bi-orthogonal transform module configured to transform the signal into a plurality of BMDCT frequency coefficients using a bi-orthogonal modified discrete cosine transform; a quantization module configured-to quantize the BMDCT frequency coefficients to produce a set of integer numbers which represents the BMDCT frequency coefficients; and a lossless encoder module configured to losslessly encode the set of integer numbers to minimize the number of bits required to represent the BMDCT frequency coefficients; and a signal analysis module configured to divide the signal into a plurality of different window types based on a characteristic of the signal and analyze the signal to determine a set of noise targets where noise can be transparently introduced into the signal to compress the signal; a stereo transform module configured to stereo transform the BMDCT signal frequency coefficients to eliminate all redundant and irrelevant bits from the BMDCT signal frequency coefficients; a rate control module configured to use the noise targets received from the signal analysis module and determine the step size required to transparently allocate noise into the BMDCT frequency coefficients to meet a required bit rate per frame; and a frame packing module configured to pack the losslessly encoded set of integer numbers, a plurality of step sizes, a plurality of transform angles, and the plurality of different window sizes into a frame to produce a formatted signal.

Claim 33

Independent Claim 233 words

An apparatus configured to encode a signal, comprising: a bi-orthogonal transform module configured to transform the signal into a plurality of BMDCT frequency coefficients using a bi-orthogonal modified discrete cosine transform; a quantization module configured-to quantize the BMDCT frequency coefficients to produce a set of integer numbers which represents the BMDCT frequency coefficients; and a lossless encoder module configured to losslessly encode the set of integer numbers to minimize the number of bits required to represent the BMDCT frequency coefficients; and a signal analysis module configured to divide the signal into a plurality of different window types based on a characteristic of the signal and analyze the signal to determine a set of noise targets where noise can be transparently introduced into the signal to compress the signal; a stereo transform module configured to stereo transform the BMDCT signal frequency coefficients to eliminate all redundant and irrelevant bits from the BMDCT signal frequency coefficients; a rate control module configured to use the noise targets received from the signal analysis module and determine the step size required to transparently allocate noise into the BMDCT frequency coefficients to meet a required bit rate per frame; and a frame packing module configured to pack the losslessly encoded set of integer numbers, a plurality of step sizes, a plurality of transform angles, and the plurality of different window sizes into a frame to produce a formatted signal.

Claim 34

Dependent Claim 23 words

The apparatus of claim 30 , wherein the different window types can be a long window, a medium window, and a short window.

Claim 35

Dependent Claim 15 words

The apparatus of claim 30 , wherein the stereo transform is a 2×2 orthogonal transform.

Claim 36

Dependent Claim 21 words

The apparatus of claim 30 , wherein the rate control module controls the quantization step sizes used by the quantization module.

Claim 37

Dependent Claim 35 words

The apparatus of claim 30 , wherein the rate control module allocates a number of bits to the current frame based on an average complexity of the stream and the complexity of the current frame.

Claim 38

Dependent Claim 24 words

The apparatus of claim 34 , wherein the complexity is measured as the number of bits which are required to code the frame transparently.

Claim 39

Dependent Claim 34 words

The apparatus of claim 30 , wherein the rate control module allows relatively flexible allocation of bits to a single frame according the following equation: tbits=avBits(0.75+0.25frameComplexity/avComplexity), where avBits is the constant bit rate target.

Claim 40

Dependent Claim 23 words

The apparatus of claim 30 , wherein the bi-orthogonal transform module provides a plurality of different analysis resolution states for different signal characteristics.

Claim 41

Dependent Claim 26 words

The apparatus of claim 37 , wherein the plurality of different analysis resolution states can be a long state, a medium state and a short state.

Claim 42

Dependent Claim 32 words

The apparatus of claim 38 , wherein the long state contains 512 signal frequency coefficients, the medium state contains 256 signal frequency coefficients, and the short state contains 128 signal frequency coefficients.

Claim 43

Dependent Claim 17 words

The apparatus of claim 39 , wherein a collection of 512 signal frequency coefficients is a sub-frame.

Claim 44

Dependent Claim 13 words

The apparatus of claim 40 , wherein there are two sub-frames per frame.

Claim 45

Dependent Claim 17 words

The apparatus of claim 41 , wherein the two sub-frames of a frame are losslessly encoded together.

Claim 46

Dependent Claim 31 words

The apparatus of claim 38 , wherein the bi-orthogonal transform module can transition directly between the long state and the medium state and between the long state and the short state.

Claim 47

Dependent Claim 20 words

The apparatus of claim 30 , wherein the quantization module quantizes the BMDCT signal frequency coefficients using pulse code modulation.

Claim 48

Dependent Claim 33 words

The apparatus of claim 30 , wherein for a given step size value, the quantized BMDCT signal frequency coefficient is represented as an integer using the following formula: C q =Round( C /step_size).

Claim 49

Dependent Claim 16 words

The apparatus of claim 45 , wherein the step size value varies across a frequency spectrum.

Claim 50

Dependent Claim 17 words

The apparatus of claim 45 , wherein the step size value is fixed within a frequency band.

Claim 51

Dependent Claim 37 words

The apparatus of claim 45 , wherein the step size value for a particular frequency band is chosen so that the noise introduced into the frequency band meets the noise target specified by the rate control module.

Claim 52

Dependent Claim 36 words

The apparatus of claim 30 , wherein the lossless encoder module uses variable length Huffman encoding and pulse code modulation in an adaptive way to minimize the number of bits required to represent BMDCT frequency coefficients.

Claim 53

Independent Claim 115 words

A method for decoding a formatted signal, comprising the steps of: decoding the coded quantized BMDCT frequency coefficients to reconstruct a quantized set of integer numbers representing BMDCT frequency coefficients of a signal, the BMDCT frequency coefficients at the decoder being represented at a reduced number of bits of precision and the BMDCT window coefficients at the encoder are configured such that the combination of forward transform and inverse transform substantially perfectly reconstruct the signal; inverse quantizing the quantized set of integer numbers to produce the BMDCT frequency coefficients of the signal; and inverse transforming the BMDCT frequency coefficients of the signal using a bi-orthogonal modified discrete cosine inverse transform module to reproduce the signal.

Claim 54

Independent Claim 115 words

A method for decoding a formatted signal, comprising the steps of: decoding the coded quantized BMDCT frequency coefficients to reconstruct a quantized set of integer numbers representing BMDCT frequency coefficients of a signal, the BMDCT frequency coefficients at the decoder being represented at a reduced number of bits of precision and the BMDCT window coefficients at the encoder are configured such that the combination of forward transform and inverse transform substantially perfectly reconstruct the signal; inverse quantizing the quantized set of integer numbers to produce the BMDCT frequency coefficients of the signal; and inverse transforming the BMDCT frequency coefficients of the signal using a bi-orthogonal modified discrete cosine inverse transform module to reproduce the signal.

Claim 55

Dependent Claim 25 words

A decoding method of claim 50 , wherein the BMDCT window coefficients at the decoder are represented at a reduced number of bits of precision.

Claim 56

Dependent Claim 50 words

The method of claim 50 , further comprising the steps of: separating the formatted bitstream into side information and quantized BMDCT frequency coefficients; and inverse stereo transforming the BMDCT frequency coefficients from a pair of channels and the transformed side information to produce the BMDCT frequency coefficients of the signal.

Claim 57

Independent Claim 129 words

An apparatus configured to decode a formatted signal, comprising: a lossless decoder module configured to losslessly decode the formatted signal to reconstruct a quantized set of integer numbers representing the BMDCT frequency coefficients of a signal; an inverse quantization module configured to inverse quantize the quantized set of integer numbers to produce the BMDCT frequency coefficients of the signal; an inverse bi-orthogonal modified discrete cosine transform module configured to inverse transform the BMDCT frequency coefficients to reproduce the signal; a frame unpacking module configured to separate the formatted bitstream into side information and quantized BMDCT frequency coefficients; and a stereo transformer configured to inverse stereo transform the BMDCT frequency coefficients from a pair of channels and the transformed side information to produce the BMDCT frequency coefficients of the signal.

Claim 58

Independent Claim 129 words

An apparatus configured to decode a formatted signal, comprising: a lossless decoder module configured to losslessly decode the formatted signal to reconstruct a quantized set of integer numbers representing the BMDCT frequency coefficients of a signal; an inverse quantization module configured to inverse quantize the quantized set of integer numbers to produce the BMDCT frequency coefficients of the signal; an inverse bi-orthogonal modified discrete cosine transform module configured to inverse transform the BMDCT frequency coefficients to reproduce the signal; a frame unpacking module configured to separate the formatted bitstream into side information and quantized BMDCT frequency coefficients; and a stereo transformer configured to inverse stereo transform the BMDCT frequency coefficients from a pair of channels and the transformed side information to produce the BMDCT frequency coefficients of the signal.

Annuity Stage	Window Opens	Payment Deadline	Surcharge Period Until	Estimated Fee	Status
3.5 Years (Year 4)	2006-10-20	2007-04-21	2007-10-21	$2000.00	Window Closed / Lapsed
7.5 Years (Year 8)	2010-10-20	2011-04-21	2011-10-21	$3760.00	Window Closed / Lapsed
11.5 Years (Year 12)	2014-10-20	2015-04-21	2015-10-21	$7700.00	Window Closed / Lapsed

System and method for noise reduction using bi-orthogonal modified discrete cosine transform

Abstract

Inventors

Bibliographic Details

Claims Analysis

References Cited (Backward Citations)

USPTO Annuity & Maintenance Windows

Assigned Categories & Tags

Assign Category Tag