Pattern recognition applying LDA and LR to optoelectronic signals of optical scanning systems

Abstract

This paper is a follow-up to previous researches already published regarding the minimization of optical-electrical noise in optical scanning systems OSS, by the implementation of computational techniques for pattern recognition generated by each reference source used to indicate a coordinate that the OSS will be monitoring. Techniques such as linear discriminant analysis LDA and linear regression LR were implemented in order to discriminate the signals caused by other sources different to the references. In order to enhance the efficiency of these models was implemented linear predictive coding LPC and quantiles as features extractors. The results were encouraging with rates of recognition greater than 91.2 %, reaching in some cases an accuracy of 100 %.

Keywords

• análisis y tratamiento de señales
• sensores e instrumentos virtuales
• ruido
• modulación
• visión 3d y estéreo