DisVAE: Disentangled Variational Autoencoder for High-Quality Facial Expression Features | IEEE Conference Publication | IEEE Xplore