American Sign Language fingerspelling recognition from video: Methods for unrestricted recognition and signer-independence

TL;DR

This paper investigates American Sign Language fingerspelling recognition from video, proposing methods for both signer-dependent and signer-independent scenarios, achieving notable error rate reductions with neural network-based models.

Contribution

It introduces segmental CRF models with deep neural features and explores signer adaptation techniques for improved recognition accuracy.

Findings

Signer-dependent error rate: up to 8%

Signer-independent error rate: up to 17%

Neural network adaptation improves signer-independent recognition

Abstract

In this thesis, we study the problem of recognizing video sequences of fingerspelled letters in American Sign Language (ASL). Fingerspelling comprises a significant but relatively understudied part of ASL, and recognizing it is challenging for a number of reasons: It involves quick, small motions that are often highly coarticulated; it exhibits significant variation between signers; and there has been a dearth of continuous fingerspelling data collected. In this work, we propose several types of recognition approaches, and explore the signer variation problem. Our best-performing models are segmental (semi-Markov) conditional random fields using deep neural network-based features. In the signer-dependent setting, our recognizers achieve up to about 8% letter error rates. The signer-independent setting is much more challenging, but with neural network adaptation we achieve up to 17% letter error rates.