This paper presents an error analysis for 3D shape estimation techniques using silhouettes from multiple views (”shape–from–silhouette”). The results of this analysis are useful for the integration of other shape reconstruction techniques into the shape– from–silhouette approach and the appropriate choice of its parameters. The analysis allows an assessment of the local 3D shape error and takes...

Choosing good error metrics for this task is challenging. We will use the geometric mean of six error metrics: two for shape, one for illumination, one for shading, one for reflectance, and the MIT intrinsic images error metric introduced in [4], which we will refer to as RS-MSE (though which the original authors call “LMSE”). We use the geometric mean of these metrics as it is insensitive to t...

The amount of training data has a crucial effect on the accuracy of HMM based meeting recognition systems. One of the largest collections of speech data is conversational telephone speech which was found to match speech in meetings well. However it is naturally recorded with limited bandwidth. In previous work we presented a scheme that allows to transform wide-band meeting data into the same s...

“Meeting” speech, for example from the RT-04S task, contains a mixture of different speaking styles that leads to word error rates higher than 25% even when close-talking microphones are being used. The problem is even more serious, as word error rates are particularly high when speakers use a clear speaking mode, for example because they want to stress an important point. Previous work showed ...

In this paper, an original spatial scene level shape error concealment technique for segmented objectbased video scenes, to be used in error-prone environments such as mobile networks, is proposed. This technique is different from existing shape concealment techniques because it considers, not only the video objects that have to be concealed, but also the context in which these objects are inse...