# US 2014/0191888 A1 · Estimator for estimating a probability distribution of a quantization index

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2014/0191888 A1
- Document ID: US20140191888A1
- Status: Published application
- Kind: A1
- Application number: 14/186,237
- Filed: 2014-02-21
- Published: 2014-07-10
- Earliest claimed: 2011-08-23
- Assignees: Huawei Technologies Co., Ltd.
- Applicants: Huawei Technologies Co., Ltd.
- Inventors: Janusz Klejsa; Minyue Li; Willem Bastiaan Kleijn
- Web page: https://patentgrep.com/patent/US20140191888A1

## CPC

- H03M 7/00: Conversion of a code where information is represented by a given sequence or number of digits to a code where the same {, similar or subset of} information is represented by a different sequence or number of digits
- H03M 7/3079: {Context modeling}
- H04N 19/463: by compressing encoding parameters before transmission
- H04N 19/91: Entropy coding, e.g. variable length coding [VLC] or arithmetic coding
- H03M: CODING; DECODING; CODE CONVERSION IN GENERAL (using fluidic means F15C4/00; optical analogue/digital converters G02F7/00; coding, decoding or code conversion, specially adapted for particular applications, see the relevant subclasses, e.g. G01D, G01R, G06F, G06T, G09G, G10L, G11B, G11C, H04B, H04L, H04M, H04N; ciphering or deciphering for cryptography or other purposes involving the need for secrecy G09C)
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

The invention relates to an estimator for estimating a probability distribution of a quantization index generated from a source coder encoding a source signal, into a sequence of quantization indices, the source signal being described by a signal model, the source coder providing a current quantization index and current side information, the estimator being configured to obtain auxiliary parameters based on a configuration of the source coder and the current available side information and the signal model, the estimator being further configured to adaptively update the probability distribution of a quantization index upon the basis of a probability density function relating to a previous state of the estimator, the auxiliary parameters, the current quantization index and the current side information.

## Claims

### Claim 1

An estimator for estimating a probability distribution of a quantization index generated from a source coder encoding a source signal into a sequence of quantization indices, the source signal being described by a signal model, the source coder providing a current quantization index and current side information, the estimator being configured to obtain auxiliary parameters based on a configuration of the source coder and the current side information and the signal model, the estimator being further configured to adaptively update the probability distribution of a quantization index upon the basis of a probability density function relating to a previous state of the estimator, the auxiliary parameters, the current quantization index and the current side information.

### Claim 2 (depends on 1)

The estimator of claim 1, wherein the current side information indicates a current configuration of the source coder, wherein the estimator is configured to derive the current auxiliary parameters from the signal model, in particular from a linear stochastic signal model of the source signal.

### Claim 3 (depends on 1)

The estimator of claim 1, wherein the estimator is configured to derive the current auxiliary parameters from the signal model and the side information, in particular obtained from a linear stochastic signal model of the source signal.

### Claim 4 (depends on 1)

The estimator of claim 1, wherein the current side information indicates a current configuration of the source coder, wherein the estimator is configured to derive the current auxiliary parameters from the signal model, in particular obtained from a linear stochastic signal model of the source signal, and wherein the estimator is further configured to update the probability density function using information on the current quantization index and the current parameters of the model of the source signal.

### Claim 5 (depends on 1)

The estimator of claim 1, wherein the estimator is configured to derive the auxiliary parameters from an Auto-Regressive-Moving-Average model of the source signal.

### Claim 6 (depends on 1)

The estimator of claim 1, wherein the side information provided by the source coder comprises a realization of a dither signal generated by the source coder.

### Claim 7 (depends on 1)

The estimator of claim 1, wherein the side information indicates an outcome of a prediction of the source signal performed in the source coder.

### Claim 8 (depends on 1)

The estimator of claim 1, wherein the side information indicates at least one of the following:
a realization of dither signal and an outcome of a closed-loop prediction of the source signal performed in the source coder; and
an open-loop prediction of the source signal performed in the source coder and delayed observation of the source signal.

### Claim 9 (depends on 1)

The estimator of claim 1, wherein the probability function relating to the state of the estimator is f(ξn|q1n−1,s1n−1)∝∫Af(ξn−1|q1n−2,s1n−1)f(ξn|ξn−1)p(qn−1|ξn−1,sn−1)dξn−1 where ξn is the n-th realization of the hidden state, A denotes the alphabet from which the hidden states take values, q1n−1 is a vector containing n−1 previous indices and s1n−1 is a vector containing n−1 previous realizations of side information S1n−1, and f(ξn−1|q1n−2,s1n−1) is the probability density function relating to the previous state of the estimator, f(ξn|ξn−1) is a conditional probability density function of the current hidden state given the previous hidden state, and p(qn−1|ξn−1,sn−1) is the probability of the previous quantization index qn−1, ξn−1 is a realization of the previous hidden state and sn−1 is the previous realization of the side information.

### Claim 10 (depends on 1)

The estimator of claim 1, being configured to adaptively update the probability distribution of a quantization index p(qn|q1n−1,s1n) upon the basis of updating the probability density function relating to the state of the estimator to in response the current side information sn, according to f(ξn|q1n−1,s1n)∝f(ξn|q1n−1,s1n−1)f(sn|ξn,q1n−1,s1n−1), where f(ξn|q1n−1,s1n−) is the probability density function relating to the previous state of the estimator and f(sn|ξn,q1n−1,s1n−1) is the probability density function of the current side information sn conditioned on a current hidden state ξn, all previous quantization indices represented by the vector q1n−1 and all the previous side information represented by s1n−1, where f(sn|ξn,q1n−1,s1n−1) is computed in response to the current side information, wherein the probability distribution of a quantization index being updated according to p(qn|q1n−1,s1n)=∫Ap(qn|ξn, sn)f(ξn|q1n−1,s1n)dξn where p(qn|ξn, sn) is the probability distribution of a quantization index qn conditioned on the current state of the estimator ξn and current side information sn.

### Claim 11 (depends on 1)

The estimator of claim 1, being configured to adaptively update the probability distribution of a quantization index upon the basis of the following estimate of the probability density function relating to the state of the estimator: f(ξn|q1n−1,s1n−1)≈N(ξn;μn,Σn), wherein the mean μn is computed as the expectation about the current hidden state Ξn (207), given all previous quantization indices and all previous realizations of side information, according to μn=E{Ξn|Q1n−1,S1n−1}, and wherein the covariance matrix is computed according to Σn=E{(Ξn−μn)(Ξn−μn)T|Q1n−1,S1n−1}, where T denotes matrix transposition and E denotes an expectation.

### Claim 12

A coding apparatus for coding a source signal, the coding apparatus comprising:
a source coder for coding the source signal into a sequence of quantization indices, the source coder configured to provide a current quantization index and current side information;
a lossless coder configured to encode the quantization indices into code words in dependency on a probability distribution of the quantization indices; and
an estimator for estimating a probability distribution of the current quantization index generated from the source coder, the source signal being described by a signal model, the estimator being configured to obtain auxiliary parameters based on a configuration of the source coder and the current side information and the signal model, the estimator being further configured to adaptively update a probability distribution of a quantization index upon the basis of a probability density function relating to a previous state of the estimator, the auxiliary parameters, the current quantization index and the current side information.

### Claim 13

A decoding apparatus for decoding a source signal being encoded by a source coder using quantization indices, the decoding apparatus comprising:
a receiver configured to receive a bit stream with code words and current side information;
an estimator for estimating a probability distribution of a current quantization index generated from the source coder, the source signal being described by a signal model, the estimator being configured to obtain auxiliary parameters based on a configuration of the source coder and current side information and the signal model, the estimator being further configured to adaptively update a probability distribution of a quantization index upon the basis of a probability density function relating to a previous state of the estimator, the auxiliary parameters, the current quantization index and the current side information; and
a decoder for decoding a sequence of quantization indices and for providing the current quantization index for a reconstruction of the source signal based on the probability distribution of a quantization index.

### Claim 14 (depends on 13)

The decoding apparatus of claim 13, wherein the current side information indicates a current configuration of the source coder, wherein the estimator is configured to derive the current auxiliary parameters from the signal model, in particular from a linear stochastic signal model of the source signal.

### Claim 15 (depends on 13)

The decoding apparatus of claim 13, wherein the estimator is configured to derive the current auxiliary parameters from the signal model and the side information, in particular obtained from a linear stochastic signal model of the source signal.

### Claim 16 (depends on 13)

The decoding apparatus of claim 13, wherein the current side information indicates a current configuration of the source coder, wherein the estimator is configured to derive the current auxiliary parameters from the signal model, in particular obtained from a linear stochastic signal model of the source signal, and wherein the estimator is further configured to update the probability density function using information on the current quantization index and the current parameters of the model of the source signal.

### Claim 17 (depends on 13)

The decoding apparatus of claim 13, wherein the estimator is configured to derive the auxiliary parameters from an Auto-Regressive-Moving-Average model of the source signal.

### Claim 18 (depends on 13)

The decoding apparatus of claim 13, wherein the side information provided by the source coder comprises a realization of a dither signal generated by the source coder.

### Claim 19 (depends on 13)

The decoding apparatus of claim 13, wherein the side information indicates an outcome of a prediction of the source signal performed in the source coder.

### Claim 20 (depends on 13)

The decoding apparatus of claim 13, wherein the side information indicates at least one of the following:
a realization of dither signal and an outcome of a closed-loop prediction of the source signal performed in the source coder; and
an open-loop prediction of the source signal performed in the source coder and delayed observation of the source signal.

### Claim 21 (depends on 13)

The decoding apparatus of claim 13, wherein the probability function relating to the state of the estimator is f(ξn|q1n−1,s1n−1)∝∫Af(ξn−1|q1n−2,s1n−1)f(ξn|ξn−1)p(qn−1|ξn−1,sn−1)dξn−1 where ξn is the n-th realization of the hidden state, A denotes the alphabet from which the hidden states take values, q1n−1 is a vector containing n−1 previous indices and s1n−1 is a vector containing n−1 previous realizations of side information S1n−1 and f(ξn−1|q1n−2,s1n−1) is the probability density function relating previous state of the estimator, f(ξn|ξn−1) is a conditional probability density function of the current hidden state given the previous hidden state, and p(qn−1|ξn−1,sn−1) is the probability of the previous quantization index qn−1,ξn−1 is a realization of the previous hidden state and sn−1 is the previous realization of the side information.

### Claim 22 (depends on 13)

The decoding apparatus of claim 13, being configured to adaptively update the probability distribution of a quantization index p(qn|q1n−1,s1n−1) upon the basis of updating the probability density function relating to the state of the estimator to in response the current side information sn according to
f(ξn|q1n−1,s1n)∝f(ξn|q1n−1,s1n−1)f(sn|ξn,q1n−1,s1n−1), where f(ξn|q1n−1,s1n−1) is the probability density function relating to the previous state of the estimator and f(sn|ξn,q1n−1,s1n−1) is the probability density function of the current side information sn, conditioned on a current hidden state ξn, all previous quantization indices represented by the vector q1n−1 and all the previous side information represented by s1n−1 where f(sn|ξn,q1n−1,s1n−1) is computed in response to the current side information, wherein the probability distribution of a quantization index being updated according to p(qn|q1n−1,s1n)=∫Ap(qn|ξn, sn)f(ξn|q1n−1,s1n)dξn where p(qn|ξn, sn) is the probability distribution of a quantization index qn conditioned on the current state of the estimator ξn and current side information sn.

### Claim 23 (depends on 13)

The decoding apparatus of claim 13, wherein the estimator is further configured to adaptively update the probability distribution of a quantization index upon the basis of the following estimate of the probability density function relating to the state of the estimator: f(ξn|q1n−1,s1n−1)≈N(ξn;μn,Σn), wherein the mean μn is computed as the expectation about the current hidden state Ξn (207), given all previous quantization indices and all previous realizations of side information, according to μn=E{Ξn|Q1n−1,S1n−1}, and wherein the covariance matrix is computed according to
Σn=E{(Ξn−μn)(Ξn−μn)T|Q1n−1,S1n−1}, where T denotes matrix transposition and E denotes an expectation.

### Claim 24

An estimation method for estimating a probability distribution of a quantization index generated from a source coder encoding a source signal into a sequence of quantization indices, the source signal being described by a signal model, the source coder providing a current quantization index and current side information, the method comprising:
obtaining auxiliary parameters based on a configuration of the source coder, the current available side information and the model of the signal model; and
adaptively updating the probability distribution of a quantization index upon the basis of a probability density function relating to a previous state of the estimator, the auxiliary parameters, the current quantization index and the current side information.

### Claim 25 (depends on 24)

A coding method for coding a source signal, the method comprising:
encoding the source signal into a sequence of quantization indices and providing a current quantization index and current side information;
losslessly encoding the quantization indices into code words in dependency on a probability distribution of the quantization indices; and
estimating the probability of a quantization index generated from the source coder according to the method for estimating a probability distribution of a quantization index of claim 24.

### Claim 26 (depends on 24)

A decoding method for decoding a source signal, the method comprising:
receiving a bit stream with code words and current side information; and
estimating the probability of a quantization index generated from the source coder according to the method for estimating a probability distribution of a quantization index of claim 24; and
decoding a sequence of quantization indices and providing a current quantization index into a reconstruction of a source signal based on the probability distribution of a quantization index.