2010
DOI: 10.1007/978-3-642-12297-2_21
|View full text |Cite
|
Sign up to set email alerts
|

Fingerspelling Recognition through Classification of Letter-to-Letter Transitions

Abstract: Abstract. We propose a new principle for recognizing fingerspelling sequences from American Sign Language (ASL). Instead of training a system to recognize the static posture for each letter from an isolated frame, we recognize the dynamic gestures corresponding to transitions between letters. This eliminates the need for an explicit temporal segmentation step, which we show is error-prone at speeds used by native signers. We present results from our system recognizing 82 different words signed by a single sign… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

1
21
0

Year Published

2012
2012
2021
2021

Publication Types

Select...
5
3
1

Relationship

0
9

Authors

Journals

citations
Cited by 24 publications
(22 citation statements)
references
References 23 publications
1
21
0
Order By: Relevance
“…For the represented mapping, our objective function predicts an entry rate of 43.9 WPM which is surprisingly close to the empirically observed rate of 40-45 WPM for experienced practitioners [30].…”
Section: Design Casessupporting
confidence: 78%
“…For the represented mapping, our objective function predicts an entry rate of 43.9 WPM which is surprisingly close to the empirically observed rate of 40-45 WPM for experienced practitioners [30].…”
Section: Design Casessupporting
confidence: 78%
“…The problem can be approached similarly to speech recognition, with letters being the analogues of words or phones, and most work thus far has indeed used HMM-based approaches with HMMs representing letters (e.g., [11,4]) or letter-to-letter transitions [12]. Most prior work has limited the vocabulary to 20-100 words, in which case it is common to obtain letter error rates of 10% or less.…”
Section: Introductionmentioning
confidence: 99%
“…A subset of ASL recognition work has focused on fingerspelling and/or handshape classification [4,27,24] and fingerspelling sequence recognition [14,20,26], where letter error rates of 10% or less have been achieved when the 1 A good deal of prior work on sign language recognition has also used other instrumentation, such as specialized gloves or depth maps [17,16,21,39,13]. These interfaces are sometimes feasible, but video remains more practical in many settings, and we restrict our discussion to video here.…”
Section: Related Workmentioning
confidence: 99%
“…Most prior work on fingerspelling recognition has assumed a closed vocabulary of fingerspelled words, often limited to 20-100 words, typically using hidden Markov models (HMMs) representing letters or letter-to-letter transitions [14,20,26]. In such settings it is common to obtain letter error rates (Levenshtein distances between hypothesized and true letter sequences, as a proportion of the number of true letters) of 10% or less.…”
Section: Introductionmentioning
confidence: 99%