Shape Descriptors for Classification of Functional Data

Curve discrimination is an important task in engineering and other sciences. We propose several shape descriptors for classifying functional data, inspired by form analysis from the image analysis field: statistical moments, coefficients of the components of independent component analysis, and two mathematical morphology descriptors (morphological covariance and spatial size distributions). These are applied to three problems: an artificial problem, a speech recognition problem, and a biomechanical application. Shape descriptors are compared with other methods in the literature, with better or similar performance obtaining.