The success of Voice-over-IP technology will crucially depend on the provision of a certain level of perceptual speech quality. In this paper we demonstrate that keeping as much corrupted speech data as possible results in a major step towards this goal. To this end, we explore the combination of two key factors which influence the speech quality as perceived by the user: networking mechanisms ...

We consider statistical models where functional data are artificially contaminated by independent Wiener processes in order to satisfy privacy constraints. show that the corrupted observations have a density which determines distribution of original random variables, masked near origin, uniquely, and we construct nonparametric estimator density. derive an upper bound for its mean integrated squ...

annihilation or reduction of each kind of noise blended in correct data signals is a field that has attracted many researchers. it is a fact that fuzzy theory presents full capability in this field. fuzzy filters are often strong in smoothing corrupted signals, whereas they have simple structures. in this paper, a new powerful yet simple fuzzy procedure is introduced for sharpness reduction in ...

Federated learning provides a communication-efficient and privacy-preserving training process by enabling statistical models with massive participants while keeping their data in local clients. However, standard federated techniques that naively minimize an average loss function are vulnerable to corruptions from outliers, systematic mislabeling, or even adversaries. In addition, it is often pr...