Single Neuron Neural Network for Low Light Photon Source Classification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for characterizing light sources at low light levels require large amounts of data and are impractical for applications like photonic technologies, metrology, imaging, and remote sensing due to the need for extensive measurements.

Innovation Solution

A method using a single neuron neural network, specifically the ADALINE model, to identify light sources by detecting individual photons, segmenting them into time bins, determining photon counts, and inputting these counts into a neural network trained on light source types, allowing for classification at extremely low light levels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional statistical methods are used to characterize light sources, then measurement accuracy is improved, but the number of measurements required increases dramatically

Engineering Contradiction:

Improvelight source characterization accuracyVSAvoidnumber of measurements required

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent transforms the characterization problem by changing from direct statistical measurement of photon arrivals to measuring time intervals between photons. This parameter transformation allows the use of machine learning algorithms that can achieve accurate light source characterization with significantly fewer measurements, resolving the contradiction between measurement precision and productivity

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces traditional statistical analysis methods with machine learning algorithms. The neural network learns to distinguish between thermal and coherent light sources by analyzing time interval patterns, substituting complex statistical computations with a trained model that requires fewer measurements while maintaining high characterization accuracy

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If machine learning methods are applied to reduce measurements, then productivity is improved, but the complexity of the detection system increases

Engineering Contradiction:

Improvenumber of measurements requiredVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the light detection process into distinct components: photon detection, time interval measurement, and machine learning classification. This segmentation allows each component to be optimized independently, managing system complexity while achieving high productivity through efficient use of machine learning on simplified time interval data

Inventive Principle:

Principle #1Segmentation

3Adaptability or versatility

If measurements are taken at extremely low light levels, then adaptability to low-light applications is improved, but measurement precision deteriorates due to statistical fluctuations

Engineering Contradiction:

Improvelow light level capabilityVSAvoidcharacterization accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent changes the measurement parameter from photon count to time interval between photons. This transformation is particularly effective at low light levels because time intervals provide sufficient statistical information even when photon counts are very low, enabling accurate light source characterization with mean photon numbers below one while maintaining measurement precision

Inventive Principle:

Principle #35Parameter changes

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach dramatically reduces the number of measurements required to identify light sources, achieving accurate classification with tens of measurements at mean photon numbers below one, outperforming traditional methods and enabling applications in LiDAR, remote sensing, and microscopy.

Implementation Method 1

detecting individual photons for a measurement time period to provide a times series of individual photon events

Methodology Applied

Scientific EffectPhotoelectric Effect: Photoelectric Effect

Data Source

PatentUS20230375399A1Smart quantum light detector

Publication Date: 2023.11.23 BOARD OF SUPERVISORS OF LOUISIANA STATE UNIV & AGRI & MECHANICAL COLLEGE

AI summary

A method and system for identification of light source types includes detecting individual photons for a measurement time period to provide a times series of individual photon events, segmenting the time series into a plurality of time bins, and determining a number of detected photons within each time bin to provide a time series of photon counts, determining a probability distribution P(n) from the time series of photon counts, the probability distribution providing the probability of detection of n photons (n=0 . . . nmax), inputting each of the values of P(n) as a nmax+1 component feature vector into a single neuron neural network that has been previously trained on a plurality of light source types, and receiving as output a classifier that has a value that identifies the light source type. An average number of photons in the plurality of time bins is less than one photon.