This paper reports recent efforts to apply the speaker-independent SPHINX-H system to the DARPA Wall Street Journal continuous speech recognition task. In SPHINX-H, we incorporated additional dynamic and speaker-normalized features, replaced discrete models with sex-dependent semi-continuous hidden Markov models, augmented within-word triphones with between-word triphones, and extended generali...

The traditional co-training algorithm, which needs a great number of unlabeled examples in advance and then trains classifiers by iterative learning approach, is not suitable for online learning of classifiers. To overcome this barrier, we propose a novel semi-supervised learning algorithm, called MAPACo-Training, by combining the co-training with the principle of Maximum A Posteriori adaptatio...

Morfessor is a family of methods for learning morphological segmentations of words based on unannotated data. We introduce a new variant of Morfessor, FlatCat, that applies a hidden Markov model structure. It builds on previous work on Morfessor, sharing model components with the popular Morfessor Baseline and Categories-MAP variants. Our experiments show that while unsupervised FlatCat does no...

This paper proposes a technique for controlling singing style in the HMM-based singing voice synthesis. A style control technique based on multiple regression HSMM (MRHSMM), which was originally proposed for the HMM-based expressive speech synthesis, is applied to the conventional technique. The idea of pitch adaptive training is introduced into the MRHSMM to improve the modeling accuracy of fu...

Department of Industrial Engineering and Management, School of Mechanical Engineering, Shanghai Jiao Tong University, 800 Dong-chuan Road, Shanghai 200240, PR China General Electric (Shanghai) Corporation, 1800 Cai Lun Road, Shanghai 201203, PR China Department of Mechanical and Industrial Engineering, 842 West Taylor Street, University of Illinois-Chicago, Chicago, IL 60607, USA d Environmenta...

For tasks such as dressing assistance, robots should be able to adapt to different user morphologies, preferences and requirements. We propose a programming by demonstration method to efficiently learn and adapt such skills. Our method encodes sensory information (relative to the human user) and motor commands (relative to the robot actuation) as a joint distribution in a hidden semi-Markov mod...

In this paper we evaluate a method for generating synthetic speech at high speaking rates based on the interpolation of hidden semi-Markov models (HSMMs) trained on speech data recorded at normal and fast speaking rates. The subjective evaluation was carried out with both blind listeners, who are used to very fast speaking rates, and sighted listeners. We show that we can achieve a better intel...

Recent brain imaging studies have provided new insight into how students are able to extend their previous problem solving skills to new but similar problems. It is still unclear, however, what the basis is of individual differences in their success at transfer. In this study, 75 subjects had been trained to solve a set of mathematical problems before they were put into the fMRI scanner, where ...

We show how to visually control acoustic speech synthesis by modelling the dependency between visual and acoustic parameters within the Hidden-Semi-Markov-Model (HSMM) based speech synthesis framework. A joint audio-visual model is trained with 3D facial marker trajectories as visual features. Since the dependencies of acoustic features on visual features are only present for certain phones, we...

In this paper we evaluate two different methods for the visual synthesis of Austrian German dialects with parametric HiddenSemi-Markov-Model (HSMM) based speech synthesis. One method uses visual dialect data, i.e. visual dialect recordings that are annotated with dialect phonetic labels, the other methods uses a standard visual model and maps dialect phones to standard phones. This second metho...