Background: Traditional algorithms for hidden Markov model decoding seek to maximize either the probability of a state path or the number of positions of a sequence assigned to th...
Abstract. Performance of real-time applications on network communication channels are strongly related to losses and temporal delays. Several studies showed that these network feat...
Pierluigi Salvo Rossi, Francesco Palmieri, Giulio ...
We compare and contrast two different models for detecting sentence-like units in continuous speech. The first approach uses hidden Markov sequence models based on N-grams and max...
Yang Liu, Andreas Stolcke, Elizabeth Shriberg, Mar...
— In this paper we propose a method that exploits 3D motion-based features between frames of 3D facial geometry sequences for dynamic facial expression recognition. An expressive...
Georgia Sandbach, Stefanos Zafeiriou, Maja Pantic,...
This paper proposes an approach to classification of adjacent segments of a time series as being either of classes. We use a hierarchical model that consists of a feature extract...