Performance Bounds of Quaternion Estimators.

Area of Science:

• Signal Processing
• Statistical Inference
• Quaternion Theory

Background:

• The widely linear (WL) estimator is a recent advancement for modeling quaternion data, encompassing both proper and improper types.
• Existing experimental results suggest WL estimators outperform strictly linear (SL) and semi-WL (SWL) estimators for improper data.
• Theoretical performance bounds for WL estimators are lacking, hindering the development of quaternion-valued learning systems for 3-D and 4-D data.

Purpose of the Study:

• To introduce a rigorous closed-form solution for quantifying the performance benefits of WL estimators.
• To analyze the mean square error (MSE) improvements offered by WL models.
• To identify conditions under which optimal WL estimation simplifies to SWL or SL estimation.

Main Methods:

• Utilizing the orthogonality principle to derive theoretical performance bounds.
• Developing a closed-form solution for performance quantification.
• Analyzing the mathematical conditions for simplification of WL estimators.

Main Results:

• A rigorous closed-form solution is presented to quantify the mean square error (MSE) benefits of WL estimators.
• The study provides a theoretical basis for the observed performance advantages of WL estimators.
• Conditions for the simplification of WL estimation to SWL or SL estimation are discussed.

Conclusions:

• The developed theoretical framework rigorously quantifies the performance gains of WL estimators in quaternion data processing.
• This work is crucial for advancing quaternion-valued learning systems in 3-D and 4-D applications.
• The findings facilitate a deeper understanding of WL estimator performance and its relationship to simpler models.