1,721,066 research outputs found
Merl -- A Mitsubishi Electric Research Laboratory
No full textworkshop web site. Matthew Brand and Ken Shan MERL---a Mitsubishi Electric Research Lab 201 Broadway, Cambridge, MA, 02139 [email protected] We introduce a method for learning a mapping between signals, and use this to drive facial animation directly from vocal cues. Instead of depending on heuristic intermediate representations such as phonemes or visemes, the system learns its own representation, which includes dynamical and contextual information. In principle, this allows the system to make optimal use of context to handle ambiguity and relatively long-lasting facial co-articulation effects. The output is a series of facial control parameters, suitable for driving many different kinds of animation ranging from photo-realistic image warps to 3D cartoon characters. 1. From lip-syncing to facial animation Psychologists and storytellers alike have observed that there is a good deal of mutual information between vocal and facial gesture [23]. Facial information can add significant
Merl -- A Mitsubishi Electric Research Laboratory
No full textWe present a Gaussain model based approach for robust and automatic extraction of roads from very low-resolution satellite imagery. First, the input image is filtered to suppress the regions that the likelihood of existing a road pixel is low. Then, the road magnitude and orientation are computed by evaluating the responses from a quadruple orthogonal line filter set. A mapping from the line domain to the vector domain is used to determine the line strength and orientation for each point. A Gaussian model is fitted to point and matching models are updated recursively
Merl A Mitsubishi Electric Research Laboratory
No full textThe most di#cult aspect of rigid-body simulation is contact modeling. Two major subproblems of contact modeling are detecting contacts and computing contact forces. Although typically studied in isolation, these problems are tightly intertwined in simulation. This paper describes a method for #tting the pieces together
Mitsubishi Electric Research Laboratories
No full textWe present a novel method, which we refer as an integral histogram, to compute the histograms of all possible target regions in an image. Our method is computationally superior and makes it possible to employ even an exhaustive search process in real-time, which was impractical before
Mitsubishi Electric Research Laboratories
No full textthis paper are: # We proposed a method to compare the variable length sequences using the HMM parameter space # We showed that the number of largest eigenvalues (in absolute value) to span subspace is proportional to the number of cluster
Mitsubishi Electric Research Laboratories
No full textThis paper presents the design of a system that interacts with users via a set of "guessers": algorithms for suggesting possible annotations to the user, in the context of learning hierarchical task model
Mitsubishi Electric Research Laboratories
No full textthis paper we will present a approach that gives accurate results through a surprisingly simple optimization process. This method is akin to the ICA approaches described in the past by Casey and Westner [1] and Smaragdis [2], it however lacks a statistical foundation and rather attempts to describe auditory scenes using a component-wise reconstruction approach. In this paper we will present two flavors of this technique, one dealing with static spectrum objects, and an extension that can deal with time-varying object
Mitsubishi Electric Research Laboratories
No full textering MERL, Mitsubishi Electric Research Labs Special features such as ridges, valleys and silhouettes, of a polygonal scene are usually displayed by explicitly identifying and then rendering `edges' for the corresponding geometry. The candidate edges are identified using the connectivity information, which requires preprocessing of the data. We present a nonobvious but surprisingly simple to implement technique to render such features without connectivity information or preprocessing. At the hardware level, based only on the vertices of a given flat polygon, we introduce new polygons, with appropriate color, shape and orientation, so that they eventually appear as special features. 1 INTRODUCTION Sharp features convey a great deal of information with very few strokes. Technical illustrations, engineering CAD diagrams as well as non-photo-realistic rendering techniques exploit these features to enhance the appearance of the underlying graphics models. The most commonly used featur
Mitsubishi Electric Research Laboratories
No full textMaking machines that can understand musical structure has long been one of the holy grails of audio processing separating overlapping sounds has been another. Here we present a simple framework initialy used for the first task, which has come to make itself very useful for source separation. We show that the same type of reasoning that allows one to find teh building elements of a musical audio stream can also be used to find and extract elements contained in auditory scenes. We relate this work with recent developments in sparse representations and dimensionality reduction and show its application in a variety of situations
Mitsubishi Electric Research Laboratories
No full textWe propose an unsupervised, low-latency traffic congestion estimation algorithm that operates on the MPEG video data. We extract congestion features directly in the compressed domain, and employ Gaussian Mixture Hidden Markov Models (GM-HMM) to detect traffic condition
- …
