Linear regression for pattern recognition
Priya Stephen, Suresh Jaganathan · 2014
This paper presents a novel method for pattern recognition problem in terms of linear regression. Normally, patterns from a single-object class lie on a linear subspace. Using this concept, we develop a linear model representing a probe image as a linear combination of class-specific galleries. Linear Regression Classification (LRC) algorithm for pattern recognition belongs to the category of nearest subspace classification. This algorithm is extensively evaluated on several standard digit and English character databases and our own Tamil character database. A comparative study with different databases and methods clearly reflects the efficiency of LRC approach for pattern recognition.