Speaker : Xuan Son Nguyen, Post-doc, Université de Caen
Title: A Neural Network Based on SPD Manifold Learning for Skeleton-based Hand Gesture Recognition
Short abstract: In this talk, we will present a neural network based on SPD manifold learning for skeleton-based hand gesture recognition. Given the stream of hand’s joint positions, the network combines two aggregation processes on respectively spatial and temporal domains. The pipeline of the network architecture consists in three main stages. The first stage is based on a convolutional layer to increase the discriminative power of learned features. The second stage relies on different architectures for spatial and temporal Gaussian aggregation of joint features. The third stage learns a final SPD matrix from skeletal data, based on a variant of stochastic gradient descent on Stiefel manifolds. We then discuss different extensions of our network for trajectory analysis and video analysis.