Spectral Classification Using Multiple Pretreatments and Classifiers

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing classification methods for optical molecular spectra, particularly in-ovo spectroscopy of chicken eggs, face challenges in achieving a balance between accuracy and robustness due to high variability and external influences, often leading to overtraining and reduced classification accuracy.

Innovation Solution

A method involving multiple classification procedures with different data pretreatments and classifiers, iteratively calculated and validated, to emphasize certain features while suppressing others, using techniques like linear discriminant analysis, neural networks, and cluster analysis to determine the probability of class membership.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single classifier is used for spectrum classification, then the classification process is simple and fast, but the accuracy and reliability are insufficient due to high variability in complex spectra

Engineering Contradiction:
Improveclassification accuracyVSAvoidclassification method complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The classification process is segmented into multiple independent classifiers, each trained on differently pre-treated spectral data. This segmentation allows each classifier to focus on specific spectral features while collectively covering the entire spectrum, thereby improving overall classification accuracy without requiring any single classifier to be overly complex

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Multiple classifiers are merged into a unified classification system where their individual results are combined through voting or probability aggregation. This merging leverages the strengths of each classifier while compensating for their individual weaknesses, achieving higher reliability than any single classifier could provide alone

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If overtraining is applied to achieve very high classification accuracy, then the classification precision improves, but the robustness dramatically decreases making the classifier sensitive to slightest deviations

Engineering Contradiction:
Improveclassification accuracyVSAvoidclassification robustness
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The system prepares multiple classifiers in advance, each trained on differently pre-treated data, to cushion against the limitations of any single classifier. This prior preparation ensures that when deviations or interference occur, other classifiers can compensate and maintain reliable classification performance

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

Solution Approach 2:

Different spectral pre-treatment methods are applied to change the parameters and characteristics of the spectral data before classification. This includes various smoothing, normalization, and transformation techniques that modify the spectral features, allowing each classifier to operate on optimally pre-processed data for its specific function

Inventive Principle:
Principle #35Parameter changes

3Reliability

If multiple data pre-treatment methods and multiple classifiers are used, then the accuracy and robustness are improved, but the computational complexity and processing time increase

Engineering Contradiction:
Improveclassification stabilityVSAvoidclassification processing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

Spectral pre-treatment is performed as a preliminary action before the actual classification process. By preprocessing the spectral data to enhance relevant features and reduce noise beforehand, the subsequent classification step becomes more efficient and requires less computational effort, balancing reliability improvement with time constraints

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3513355B1Method for classifying spectra of objects having complex information content
Publication Date: 2026.04.29 TECHNISCHE UNIVERSITAT DRESDEN
  • EP3513355B1 patent drawingFigure 1
  • EP3513355B1 patent drawingFigure 2a~2d
  • EP3513355B1 patent drawingFigure 3a~3d

AI summary

The invention relates to a method for classifying spectra of objects having complex information content after registration of the spectra by using a method for pre-processing data and by using a method, associated with the data pre-processing, for classification with the calculation of a classifier. After the registration of the spectra and the pre-processing of the spectra, a multiple classification method is performed, having at least two different methods of the data pre-processing of the spectra and of the method, associated with the particular data-preprocessing, for classification. After the registration and the data pre-processing of the spectra, the following steps are performed: calculating a plurality of classifiers of the series per type of data pre-processing, determining the classifiers of the series with iterative adjustment and validation, calculating probabilities of the class association, all classifiers of the series or classifiers being equally incorporated in the determination of a classification result.