# US 2026/0212177 A1 · Constellation diagram determination method and apparatus, and device and storage medium

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

- Number: US 2026/0212177 A1
- Document ID: US20260212177A1
- Status: Published application
- Kind: A1
- Application number: 19/140,096
- Filed: 2024-01-25
- Published: 2026-07-23
- Earliest claimed: 2023-05-08
- Assignees: Zte Corporation
- Applicants: Zte Corporation
- Inventors: Wanjin ZHANG; Zhen YANG
- Web page: https://patentgrep.com/patent/US20260212177A1

## CPC

- G06N 3/08: Learning methods
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/048: Activation functions
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/09: Supervised learning
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- H04L 25/02: Details {; arrangements for supplying electrical power along data transmission lines (systems for transmitting signals via power distribution lines H04B3/54)}
- H04L 25/0254: {using neural network algorithms}
- H04L 25/067: {providing soft decisions, i.e. decisions together with an estimate of reliability (H04L25/068 and H04L25/069 take precedence; sequence estimation techniques H04L25/03178)}
- H04L 27/34: Amplitude- and phase-modulated carrier systems, e.g. quadrature-amplitude modulated carrier systems
- H04L 27/38: Demodulator circuits; Receiver circuits
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Provided are a constellation diagram determination method, a device, and a storage medium. The constellation diagram determination method includes: acquiring channel characteristic information and M modulation bit sequences; and determining a constellation diagram corresponding to the M modulation bit sequences through a pre-trained neural network, where M denotes the number of points in the modulation constellation diagram, and the constellation diagram corresponds to the channel characteristic information.

## Claims

### Claim 1

A constellation diagram determination method, comprising:
acquiring channel characteristic information and M modulation bit sequences; and
determining a constellation diagram corresponding to the M modulation bit sequences through a pre-trained neural network,
wherein M denotes a number of points in the constellation diagram, and the constellation diagram corresponds to the channel characteristic information.

### Claim 2 (depends on 1)

The constellation diagram determination method according to claim 1, wherein the neural network comprises a modulation subnetwork and a demodulation subnetwork; and
the neural network is determined in the following manner:
constructing the modulation subnetwork according to a modulation manner;
constructing the demodulation subnetwork according to a demodulation manner;
initializing a network parameter of the modulation subnetwork according to the modulation manner;
initializing a network parameter of the demodulation subnetwork according to the demodulation manner; and
training the initialized modulation subnetwork and the initialized demodulation subnetwork according to a training data set to obtain a trained neural network,
wherein the training data set comprises a plurality of pieces of sample channel characteristic information and demodulation bit probabilities corresponding to the M modulation bit sequences under the plurality of pieces of sample channel characteristic information.

### Claim 3 (depends on 2)

The constellation diagram determination method according to claim 2, wherein the modulation subnetwork comprises:
a bit selection unit configured to select a first bit and a second bit from a modulation bit sequence of the M modulation bit sequences;
a coordinate calculation unit configured to determine absolute values of coordinates of a constellation point corresponding to the modulation bit sequence through the first bit;
a sign calculation unit configured to determine, through the second bit, a quadrant in which the constellation point corresponding to the modulation bit sequence is located; and
a normalization unit configured to normalize the coordinates of the constellation point corresponding to the modulation bit sequences.

### Claim 4 (depends on 3)

The constellation diagram determination method according to claim 3, wherein N bit selection units are provided, and N is equal to log2(M)/2; and
Q coordinate calculation units are provided, and Q is equal to log2(M)/2−1.

### Claim 5 (depends on 3)

The constellation diagram determination method according to claim 3, wherein the modulation subnetwork further comprises:
a coordinate transformation unit configured to transform the coordinates of the constellation point corresponding to the modulation bit sequence from a first coordinate system to a second coordinate system.

### Claim 6 (depends on 3)

The constellation diagram determination method according to claim 3, wherein the modulation subnetwork further comprises:
a first identity unit configured to increase a network parameter amount of the modulation subnetwork.

### Claim 7 (depends on 2)

The constellation diagram determination method according to claim 2, wherein the demodulation subnetwork comprises:
a log-likelihood ratio (LLR) calculation unit configured to determine an LLR of a received signal; and
a bit probability calculation unit configured to determine a demodulation bit probability through the LLR of the received signal.

### Claim 8 (depends on 7)

The constellation diagram determination method according to claim 7, wherein P LLR calculation units are provided, and P is equal to log2(M)/2.

### Claim 9 (depends on 7)

The constellation diagram determination method according to claim 7, wherein the demodulation subnetwork further comprises:
a coordinate inverse transformation unit configured to transform coordinates of the received signal from a second coordinate system to a first coordinate system.

### Claim 10 (depends on 7)

The constellation diagram determination method according to claim 7, wherein the demodulation subnetwork further comprises:
a second identity unit configured to increase a network parameter amount of the demodulation subnetwork.

### Claim 11 (depends on 3)

The constellation diagram determination method according to claim 3, wherein initializing the network parameter of the modulation subnetwork according to the modulation manner comprises:
determining a first initial parameter of the modulation subnetwork according to a calculation formula of the modulation manner; and
initializing the modulation subnetwork according to the first initial parameter.

### Claim 12 (depends on 11)

The constellation diagram determination method according to claim 11, wherein initializing the modulation subnetwork according to the first initial parameter comprises:
scrambling the first initial parameter and initializing the modulation subnetwork according to the scrambled first initial parameter.

### Claim 13 (depends on 11)

The constellation diagram determination method according to claim 11, further comprising:
controlling a training attribute of the bit selection unit and a training attribute of the sign calculation unit to be untrainable; or
controlling a training attribute of the bit selection unit and a training attribute of the sign calculation unit to be trainable.

### Claim 14 (depends on 2)

The constellation diagram determination method according to claim 2, wherein initializing the network parameter of the demodulation subnetwork according to the demodulation manner comprises:
determining a second initial parameter of the demodulation subnetwork according to a calculation formula of the demodulation manner; and
initializing the demodulation subnetwork according to the second initial parameter.

### Claim 15 (depends on 14)

The constellation diagram determination method according to claim 14, wherein initializing the demodulation subnetwork according to the second initial parameter comprises:
scrambling the second initial parameter and initializing the demodulation subnetwork according to the scrambled second initial parameter.

### Claim 16

(canceled)

### Claim 17

An electronic device, comprising a memory and a processor, wherein the memory is configured to store a computer program that, when executed by the processor, causes the processor to perform the following:
acquiring channel characteristic information and M modulation bit sequences; and
determining a constellation diagram corresponding to the M modulation bit sequences through a pre-trained neural network,
wherein M denotes a number of points in the constellation diagram, and the constellation diagram corresponds to the channel characteristic information.

### Claim 18

A non-transitory storage medium storing a computer program that, when executed by a processor, causes the processor to perform the following:
acquiring channel characteristic information and M modulation bit sequences; and
determining a constellation diagram corresponding to the M modulation bit sequences through a pre-trained neural network,
wherein M denotes a number of points in the constellation diagram, and the constellation diagram corresponds to the channel characteristic information.

### Claim 19 (depends on 17)

The electronic device according to claim 17, wherein the neural network comprises a modulation subnetwork and a demodulation subnetwork; and
the neural network is determined in the following manner:
constructing the modulation subnetwork according to a modulation manner;
constructing the demodulation subnetwork according to a demodulation manner;
initializing a network parameter of the modulation subnetwork according to the modulation manner;
initializing a network parameter of the demodulation subnetwork according to the demodulation manner; and
training the initialized modulation subnetwork and the initialized demodulation subnetwork according to a training data set to obtain a trained neural network,
wherein the training data set comprises a plurality of pieces of sample channel characteristic information and demodulation bit probabilities corresponding to the M modulation bit sequences under the plurality of pieces of sample channel characteristic information.

### Claim 20 (depends on 19)

The electronic device according to claim 19, wherein the modulation subnetwork comprises:
a bit selection unit configured to select a first bit and a second bit from a modulation bit sequence of the M modulation bit sequences;
a coordinate calculation unit configured to determine absolute values of coordinates of a constellation point corresponding to the modulation bit sequence through the first bit;
a sign calculation unit configured to determine, through the second bit, a quadrant in which the constellation point corresponding to the modulation bit sequence is located; and
a normalization unit configured to normalize the coordinates of the constellation point corresponding to the modulation bit sequences.

### Claim 21 (depends on 20)

The electronic device according to claim 20, wherein N bit selection units are provided, and N is equal to log2(M)/2; and
Q coordinate calculation units are provided, and Q is equal to log2(M)/2−1.
