A wearable computer based American sign language recognizer

Starner﻿﻿, Thad; Weaver, Joshua; Pentland, Alex

doi:10.1007/bfb0055972

Cited by 29 publications

(29 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(For conventional BaumWelch re-estimation with a uniform prior, one simply setsP ijj = j;i = P i j;i .) We compared entropically and conventionally estimated continuous-output HMMs on sign-language gesture data provided by a computer vision lab [Starner and Pentland, 1997]. Experimental conditions for this and all subsequent tests are detailed in appendix B. Entropic estimation consistently yielded HMMs with simpler transition matrices having many parameters at or near zero (e.g., figure 3)-lower-entropy dynamical models.…”

Section: Continuous-output Hmmsmentioning

confidence: 99%

See 1 more Smart Citation

Structure Learning in Conditional Probability Models via an Entropic Prior and Parameter Extinction

Brand

1999

Neural Computation

126

View full text Add to dashboard Cite

We introduce an entropic prior for multinomial parameter estimation problems and solve for its maximum a posteriori (MAP) estimator. The prior is a bias for maximally structured and minimally ambiguous models. In conditional probability models with hidden state, iterative MAP estimation drives weakly supported parameters toward extinction, effectively turning them off. Thus structure discovery is folded into parameter estimation. We then establish criteria for simplifying a probabilistic model's graphical structure by trimming parameters and states, with a guarantee that any such deletion will increase the posterior probability of the model. Trimming accelerates learning by sparsifying the model. All operations monotonically and maximally increase the posterior probability, yielding structure-learning algorithms only slightly slower than parameter estimation via expectation-maximization (EM), and orders of magnitude faster than search-based structure induction. When applied to hidden Markov model (HMM) training, the resulting models show superior generalization to held-out test data. In many cases the resulting models are so sparse and concise that they are interpretable, with hidden states that strongly correlate with meaningful categories. Neural ComputationThis work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved. AbstractWe introduce an entropic prior for multinomial parameter estimation problems and solve for its maximum a posteriori (MAP) estimator. The prior is a bias for maximally structured and minimally ambiguous models. In conditional probability models with hidden state, iterative MAP estimation drives weakly supported parameters toward extinction, effectively turning them off. Thus structure discovery is folded into parameter estimation. We then establish criteria for simplifying a probabilistic model's graphical structure by trimming parameters and states, with a guarantee that any such deletion will increase the posterior probability of the model. Trimming accelerates learning by sparsifying the model. All operations monotonically and maximally increase the posterior probability, yielding structure-learning algorithms only slightly slower than parameter estimation via expectation-maximization (EM), and orders of magnitude faster than search-based structure induction. When applied to hidden Markov model (HMM) training, the resulting models show superior generalization to held-ou...

show abstract

Section: Continuous-output Hmmsmentioning

confidence: 99%

“…Thad Starner provided the computer vision sign language data, also used in [Starner and Pentland, 1997]. Many thanks to local and anonymous reviewers for pointing out numerous avenues of improvement.…”

Section: Acknowledgmentsmentioning

confidence: 99%

Structure Learning in Conditional Probability Models via an Entropic Prior and Parameter Extinction

Brand

1999

Neural Computation

126

View full text Add to dashboard Cite

show abstract

“…As with mobile computing, a wearable computer user's context is constantly changing. Applications in this area have ranged from understanding sign language [25] to tour guides [26] to airplane maintenance [27]. In all these applications, context has been used to improve the user's experience.…”

Section: Background and Related Workmentioning

confidence: 99%

CyberDesk: a framework for providing self-integrating context-aware services

Dey¹,

Abowd²,

Wood³

1998

Knowledge-Based Systems

View full text Add to dashboard Cite

Applications are often designed to take advantage of the potential for integration with each other via shared information. Current approaches for integration are limited, affecting both the programmer and end-user. In this paper, we present CyberDesk, a framework for self-integrating software in which integration is driven by user context. It relieves the burden on programmers by removing the necessity to predict how software should be integrated. It also relieves the burden from users by removing the need to understand how to make different software components work together. ᭧

show abstract

“…The resulting character recognition system can be also used in mobile communication and computing devices such as mobile phones, laptop computers, handheld computers, and PDAs. The advantages of our computer vision based text entry system compared to other vision based systems [10][11][12] algorithms exist for the recognition of unistroke characters and they can be implemented in real time with very high recognition accuracy. After a few minutes of studying the Graffiti alphabet, about 86% accuracy is possible.…”

Section: Introductionmentioning

confidence: 99%

Vision-based single-stroke character recognition for wearable computing

et al. 2001

View full text Add to dashboard Cite

We describe a method for recognizing the regular characters drawn by hand gestures or by a pointer on the forearm of the user captured by a head mounted camera for wearable computing. We assume that each character is drawn by a single stroke and in an isolated manner as in Graffiti. Recognition is performed by a bank of finite state machines whose input is the chain code of the hand drawn character.

show abstract

A wearable computer based American sign language recognizer

Abstract: Abstract

Cited by 29 publications

References 14 publications

Structure Learning in Conditional Probability Models via an Entropic Prior and Parameter Extinction

Structure Learning in Conditional Probability Models via an Entropic Prior and Parameter Extinction

CyberDesk: a framework for providing self-integrating context-aware services

Vision-based single-stroke character recognition for wearable computing

Contact Info

Product

Resources

About