Sign language is a visual language that is commonly used by people with hearing loss as their primary form of communication. However, many individuals with hearing loss do not use sign language and often experience social isolation as a result. To address this issue, it is important to develop human-computer interface systems that can provide a social platform for the hearing-impaired community. Currently available commercial sign language translation systems are sensor-based, expensive, and difficult to use. While vision-based systems are needed, they face several challenges that need to be overcome. Previous techniques for continuous sign language recognition have used hidden Markov models (HMM), which have limitations in capturing temporal information. To overcome these limitations, various machine learning approaches are being applied to convert hand and sign language motions into spoken or written language. This study focuses on comparing different deep learning techniques for recognizing sign language. The survey conducted aims to provide a comprehensive overview of the most recent approaches and challenges in the field of sign language recognition. The authors analyze recent developments in deep learning-based sign language recognition (SLR) and discuss techniques such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and long short-term memory (LSTM) networks. The study also explores feature extraction methods and evaluates the performance of different algorithms for SLR. The authors emphasize the importance of developing accurate and efficient SLR systems that can enhance communication for individuals with hearing loss. Overall, this research contributes to the advancement of human-computer interaction technology by providing insights into the current state of sign language recognition techniques and identifying areas for further improvement.
- Sign language is a visual language used by people with hearing loss
- Many individuals with hearing loss do not use sign language and experience social isolation
- Developing human-computer interface systems for the hearing-impaired community is important
- Current commercial sign language translation systems are sensor-based, expensive, and difficult to use
- Vision-based systems face challenges that need to be overcome
- Previous techniques for continuous sign language recognition have limitations in capturing temporal information
- Machine learning approaches are being applied to convert hand and sign language motions into spoken or written language
- This study compares different deep learning techniques for recognizing sign language
- The survey provides an overview of recent approaches and challenges in sign language recognition
- Techniques such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and long short-term memory (LSTM) networks are discussed
- Feature extraction methods are explored and algorithm performance is evaluated for sign language recognition
- Accurate and efficient SLR systems can enhance communication for individuals with hearing loss
- The research contributes to the advancement of human-computer interaction technology
Sign language is a way of communicating using hand motions for people who can't hear well. Some people with hearing loss feel lonely because they don't use sign language. It's important to make computer systems that help people with hearing loss. The current sign language translation systems are expensive and hard to use. There are challenges in making computer systems that understand sign language well. This study compares different ways of teaching computers to understand sign language. It explores methods like CNNs, RNNs, and LSTM networks. The research helps improve technology that helps people with hearing loss communicate better."
Definitions- Sign language: A visual language that uses hand motions to communicate.
- Hearing loss: When someone can't hear well or at all.
- Human-computer interface: Technology that allows humans to interact with computers.
- Sensor-based: Using sensors or devices to gather information.
- Vision-based: Using cameras or visual information for a system to work.
- Continuous sign language recognition: Teaching computers to understand ongoing sign language conversations.
- Machine learning: Teaching computers how to learn and make decisions on their own.
- Spoken or written language: Words spoken out loud or written down as text.
- Convolutional neural networks (CNNs): A type of machine learning model used for image recognition tasks.
- Recurrent neural networks (RNNs): A type of machine learning model used for sequence data, like speech or text.
- Long short-term memory (LSTM) networks: A specific type of R
Sign Language Recognition: A Comprehensive Overview of Deep Learning Techniques
Sign language is a visual language that is commonly used by people with hearing loss as their primary form of communication. Unfortunately, many individuals with hearing loss do not use sign language and often experience social isolation as a result. To address this issue, it is important to develop human-computer interface systems that can provide a social platform for the hearing-impaired community. Currently available commercial sign language translation systems are sensor-based, expensive, and difficult to use. While vision-based systems are needed, they face several challenges that need to be overcome.
In order to facilitate communication between individuals who use sign language and those who don’t understand it, machine learning approaches have been developed to convert hand and sign language motions into spoken or written language. This study focuses on comparing different deep learning techniques for recognizing sign language in order to provide a comprehensive overview of the most recent approaches and challenges in the field of sign language recognition (SLR).
Previous Approaches
Previous techniques for continuous SLR have used hidden Markov models (HMM), which have limitations in capturing temporal information. In contrast, deep learning methods such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and long short-term memory (LSTM) networks offer more accurate results due to their ability to capture temporal information from video sequences.
Feature Extraction Methods
The authors also explore feature extraction methods for SLR using CNNs and RNNs. Feature extraction involves extracting features from an image or video sequence that can be used by the algorithm for classification purposes. The authors discuss two main types of feature extraction methods: global feature extraction using CNNs and local feature extraction using RNNs. Global feature extraction extracts features from an entire image or video sequence while local feature extraction extracts features from individual frames within a sequence.
Evaluating Performance
The authors evaluate the performance of various algorithms for SLR including CNNs, RNNs, LSTM networks, HMMs, support vector machines (SVMs), random forests (RFs), k nearest neighbors (KNN) algorithms etc., based on accuracy metrics such as precision recall curves; F1 scores; confusion matrices; receiver operating characteristic curves; mean average precision scores; area under curve values etc., They also compare these algorithms based on computational complexity parameters such as training time; inference time etc.,
Conclusion
The authors emphasize the importance of developing accurate and efficient SLR systems that can enhance communication for individuals with hearing loss through improved user interfaces and increased accessibility options . Overall , this research contributes to the advancement of human - computer interaction technology by providing insights into current state -of -the -art techniques in SLR , identifying areas where further improvements could be made ,and offering guidance on how best to approach future developments .
