Kernelized Covariance for Action Recognition

TL;DR

This paper introduces Kernelized-COV, a novel method that extends covariance matrices with kernel functions to better capture non-linear relationships in data, improving action recognition performance.

Contribution

It presents a mathematically rigorous pipeline to kernelize covariance matrices, enhancing their descriptive power for complex data relationships in action recognition.

Findings

Outperforms previous methods on benchmark datasets

Scores on par or better than state-of-the-art approaches

Validates the effectiveness of kernelized covariance in capturing non-linear dependencies

Abstract

In this paper we aim at increasing the descriptive power of the covariance matrix, limited in capturing linear mutual dependencies between variables only. We present a rigorous and principled mathematical pipeline to recover the kernel trick for computing the covariance matrix, enhancing it to model more complex, non-linear relationships conveyed by the raw data. To this end, we propose Kernelized-COV, which generalizes the original covariance representation without compromising the efficiency of the computation. In the experiments, we validate the proposed framework against many previous approaches in the literature, scoring on par or superior with respect to the state of the art on benchmark datasets for 3D action recognition.