Speech Emotion Recognition (SER) is an important and challenging task for human-computer interaction. In the literature deep learning architectures have been shown to yield state-ofthe-art performance on this task when the model is trained and evaluated on the same corpus. However, prior work has indicated that such systems often yield poor performance on unseen data. To improve the generalisation capabilities of emotion recognition systems one possible approach is cross-corpus training, which consists of training the model on an aggregation of different corpora. In this paper we present an analysis of the generalisation capability of deep learning models using crosscorpus training with six different speech emotion corpora. We evaluate the models on an unseen corpus and analyse the learned representations using the t-SNE algorithm, showing that architectures based on recurrent neural networks are prone to overfit the corpora present in the training set, while architectures based on convolutional neural networks (CNNs) show better generalisation capabilities. These findings indicate that (1) cross-corpus training is a promising approach for improving generalisation and (2) CNNs should be the architecture of choice for this approach.

The woman pelvic system involves multiple organs, muscles, ligaments, and fasciae where different pathologies may occur. Here we are most interested in abnormal mobility, often caused by complex and not fully understood mechanisms. Computer simulation and modeling using the finite element (FE) method are the tools helping to better understand the pathological mobility, but of course patient-specific models are required to make contribution to patient care. These models require a good representation of the pelvic system geometry, information on the material properties, boundary conditions and loading. In this contribution we focus on the relative influence of the inaccuracies in geometry description and of uncertainty of patient-specific material properties of soft connective tissues. We conducted a comparative study using several constitutive behavior laws and variations in geometry description resulting from the imprecision of clinical imaging and image analysis. We find that geometry seems to have the dominant effect on the pelvic organ mobility simulation results. Provided that proper finite deformation non-linear FE solution procedures are used, the influence of the functional form of the constitutive law might be for practical purposes negligible. These last findings confirm similar results from the fields of modeling neurosurgery and abdominal aortic aneurysms.

No abstract