Machine learning based sign language recognition: a review and its research frontier
- PDF / 911,134 Bytes
- 20 Pages / 595.276 x 790.866 pts Page_size
- 44 Downloads / 196 Views
ORIGINAL RESEARCH
Machine learning based sign language recognition: a review and its research frontier R. Elakkiya1 Received: 23 April 2020 / Accepted: 23 July 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract In the recent past, research in the field of automatic sign language recognition using machine learning methods have demonstrated remarkable success and made momentous progression. This research article investigates the impact of machine learning in the state of the art literature on sign language recognition and classification. It highlights the issues faced by the present recognition system for which the research frontier on sign language recognition intends the solutions. In this article, around 240 different approaches have been compared that explore sign language recognition for recognizing multilingual signs. The research done by various authors is also studied, and some of the important research articles are also discussed in this article. Based on the inferences from these approaches, this article discussed how machine learning methods could benefit the field of automatic sign language recognition and the potential gaps that machine learning approaches need to address for the real-time sign language recognition. Keywords Sign language recognition · Subunit framework · Feature extraction · Movement epenthesis · Machine learning
1 Introduction The present state of the art on Sign Language Recognition (SLR) is approximately 30 years behind speech recognition systems owing manifold causes. One of the primary reasons behind this is processing and recognizing the two-dimensional video signals are highly complex than processing the one-dimensional audio signals. Besides, sign language lexical and semantic items are not yet fully discovered, and also, no standard dictionaries exist. Apart from these, for such large number of signs no common definitions exist. Sign language recognition and classification has reaches its perseverance of research publications in the beginnings of 1990s. Mostly, the presented system takes almost of 10 s to process the signer’s video and translate to text. The data acquisition methods plays vital role in classifying the primary features of different works on SLR. Many researchers have used data gloves or cyber gloves to extract the features of the manual and non-manual components of * R. Elakkiya [email protected] 1
Centre for Information Super Highway (CISH), School of Computing, SASTRA Deemed To Be University, Thanjavur 613401, India
the signs because of the reliability of the sensor-based SLR systems. However, for the signer, the use of such sensors are quiet unnatural and more restrictive. Also, the practical implementations of sensor-based SLR systems are infeasible due to the expensive nature of sensors. On the other hand, vision-based SLR systems (Elakkiya and Selvamani 2015; Selvamani and Elakkiya 2017) have greatly influenced the researchers by their heftiness and their ability to handle cluttered, dynamic inhomogeneous environments
Data Loading...
