Taste Recognition in E-Tongue Using Local Discriminant Preservation Projection

Area of Science:

• Chemometrics
• Sensor Technology
• Machine Learning

Background:

• Electronic tongue (E-Tongue) systems offer a novel approach to taste analysis.
• Conventional E-Tongue data often presents inseparable patterns due to system noise and environmental variations.
• Advanced algorithms are needed to improve the accuracy and reliability of taste recognition.

Purpose of the Study:

• To develop and evaluate a voltammetric E-Tongue system for taste recognition.
• To propose and validate the Local Discriminant Preservation Projection (LDPP) model for enhancing data separability.
• To assess the performance of LDPP combined with various classifiers for taste identification.

Main Methods:

• Construction of a voltammetric E-Tongue system measuring 13 diverse liquid samples.
• Application of the Local Discriminant Preservation Projection (LDPP) subspace learning algorithm.
• Utilized Support Vector Machine (SVM), Extreme Learning Machine (ELM), and Kernelized ELM (KELM) as classifiers.

Main Results:

• The proposed E-Tongue system demonstrated effectiveness in recognizing multiple tastes.
• The LDPP model showed higher tolerance to outliers and improved data separability compared to conventional methods.
• The LDPP-based KELM classifier achieved a peak taste recognition performance of 98%.

Conclusions:

• The developed voltammetric E-Tongue integrated with the LDPP-based KELM classifier is highly effective for taste recognition.
• The LDPP model enhances the robustness and accuracy of E-Tongue data analysis.
• The study provides a benchmark dataset and codes for future research in E-Tongue technology.