SNN Signal Processing for Low-Bit Quantization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current wireless communication systems face challenges in energy efficiency and data transmission capacity, particularly with the increasing demand for higher data rates and low-latency services in next-generation mobile communication systems.

Innovation Solution

A method and device for transmitting and receiving signals using low-bit quantization based on differential phase shift keying (DPSK) modulation and demodulation, employing a spike neural network (SNN) for pre-processing and post-processing to improve energy efficiency and understand non-linear relations between transmission and reception signals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If low-bit quantization is used to reduce power consumption, then energy efficiency is improved, but reception performance deteriorates due to quantization errors

Engineering Contradiction:

Improvepower consumptionVSAvoidreception performance

Core Design Contradiction:

Use of energy by moving objectVSReliability

Solution Approach 1:

The patent applies preliminary action by pre-training a neural network model to learn the inverse characteristics of the quantization channel before actual signal transmission. This pre-trained model is then used to compensate for quantization errors during reception, thereby maintaining reliable signal detection even with low-bit quantization that reduces power consumption.

Inventive Principle:

Principle #10Preliminary action

2Productivity

If high data transmission rates are implemented to meet increasing data traffic demand, then productivity is improved, but energy consumption increases

Engineering Contradiction:

Improvedata transmission rateVSAvoidenergy consumption

Core Design Contradiction:

ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent changes the quantization bit depth parameter to achieve low-bit quantization (1-3 bits), which significantly reduces the energy consumption of the ADC while maintaining acceptable data transmission rates. Combined with neural network-based compensation, this parameter change enables energy-efficient high-speed data transmission.

Inventive Principle:

Principle #35Parameter changes

3Use of energy by moving object

If low-bit quantization (1-3 bits) is applied to ADC to save energy, then use of energy is improved, but measurement precision deteriorates

Engineering Contradiction:

ImproveADC energy consumptionVSAvoidsignal quantization accuracy

Core Design Contradiction:

Use of energy by moving objectVSMeasurement precision

Solution Approach 1:

The patent introduces a neural network model as an intermediary between the low-bit quantized signal and the final decoded signal. This neural network learns the complex non-linear relationship introduced by low-bit quantization and compensates for the measurement precision loss, enabling accurate signal recovery despite coarse quantization.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12003277B2Method for transmitting or receiving signal in low-bit quantization system and device therefor

Publication Date: 2024.06.04 LG ELECTRONICS INC

AI summary

The present disclosure relates to a method for transmitting or receiving a signal by a reception device in a wireless communication system, and the method may comprise the steps of: receiving, from a transmission device, a signal modulated on the basis of a differential phase shift keying (DPSK) scheme; converting the received signal into an input signal of a spiking neural network (SNN); calculating an output value through the spiking neural network previously learned; and converting the output value into an input signal of a channel decoder.