American Sign Language Words Recognition of Skeletal Videos Using Processed Video Driven Multi-Stacked Deep LSTM

Intelligent real-life key-pixel image detection system for early Arabic sign language learners

Lack of an effective early sign language learning framework for a hard-of-hearing population can have traumatic consequences, causing social isolation and unfair treatment in workplaces. Alphabet and digit detection methods have been the basic framework for early sign language learning but are restricted by performance and accuracy, making it difficult to detect signs in real life. This article proposes an improved sign language detection method for early sign language learners based on the You Only Look Once version 8.0 (YOLOv8) algorithm, referred to as the intelligent sign language detection system (iSDS), which exploits the power of deep learning to detect sign language-distinct features. The iSDS method could overcome the false positive rates and improve the accuracy as well as the speed of sign language detection. The proposed iSDS framework for early sign language learners consists of three basic steps: (i) image pixel processing to extract features that are underrepresented in the frame, (ii) inter-dependence pixel-based feature extraction using YOLOv8, (iii) web-based signer independence validation. The proposed iSDS enables faster response times and reduces misinterpretation and inference delay time. The iSDS achieved state-of-the-art performance of over 97% for precision, recall, and F1-score with the best mAP of 87%. The proposed iSDS method has several potential applications, including continuous sign language detection systems and intelligent web-based sign recognition systems.

Multi-cue temporal modeling for skeleton-based sign language recognition

Sign languages are visual languages used as the primary communication medium for the Deaf community. The signs comprise manual and non-manual articulators such as hand shapes, upper body movement, and facial expressions. Sign Language Recognition (SLR) aims to learn spatial and temporal representations from the videos of the signs. Most SLR studies focus on manual features often extracted from the shape of the dominant hand or the entire frame. However, facial expressions combined with hand and body gestures may also play a significant role in discriminating the context represented in the sign videos. In this study, we propose an isolated SLR framework based on Spatial-Temporal Graph Convolutional Networks (ST-GCNs) and Multi-Cue Long Short-Term Memorys (MC-LSTMs) to exploit multi-articulatory (e.g., body, hands, and face) information for recognizing sign glosses. We train an ST-GCN model for learning representations from the upper body and hands. Meanwhile, spatial embeddings of hand shape and facial expression cues are extracted from Convolutional Neural Networks (CNNs) pre-trained on large-scale hand and facial expression datasets. Thus, the proposed framework coupling ST-GCNs with MC-LSTMs for multi-articulatory temporal modeling can provide insights into the contribution of each visual Sign Language (SL) cue to recognition performance. To evaluate the proposed framework, we conducted extensive analyzes on two Turkish SL benchmark datasets with different linguistic properties, BosphorusSign22k and AUTSL. While we obtained comparable recognition performance with the skeleton-based state-of-the-art, we observe that incorporating multiple visual SL cues improves the recognition performance, especially in certain sign classes where multi-cue information is vital. The code is available at: https://github.com/ogulcanozdemir/multicue-slr.

Application of Wearable Gloves for Assisted Learning of Sign Language Using Artificial Neural Networks

This study proposes the design and application of wearable gloves that can recognize sign language expressions from input images via long short-term memory (LSTM) network models and can learn sign language through finger movement generation and vibration motor feedback. It is difficult for nondisabled people who do not know sign language to express sign language accurately. Therefore, we suggest the use of wearable gloves for sign language education to help nondisabled people learn and accurately express sign language. The wearable glove consists of a direct current motor, a link (finger exoskeleton) that can generate finger movements, and a flexible sensor that recognizes the degree of finger bending. When the coordinates of the hand move in the input image, the sign language motion is fed back through the vibration motor attached to the wrist. The proposed wearable glove can learn 20 Korean sign language words, and the data used for learning are configured to represent the joint coordinates and joint angles of both the hands and body for these 20 sign language words. Prototypes were produced based on the design, and it was confirmed that the angle of each finger could be adjusted. Through experiments, a sign language recognition model was selected, and the validity of the proposed method was confirmed by comparing the generated learning results with the data sequence. Finally, we compared and verified the accuracy and learning loss using a recurrent neural network and confirmed that the test results of the LSTM model showed an accuracy of 85%.

Lie Recognition with Multi-Modal Spatial–Temporal State Transition Patterns Based on Hybrid Convolutional Neural Network–Bidirectional Long Short-Term Memory

Recognition of lying is a more complex cognitive process than truth-telling because of the presence of involuntary cognitive cues that are useful to lie recognition. Researchers have proposed different approaches in the literature to solve the problem of lie recognition from either handcrafted and/or automatic lie features during court trials and police interrogations. Unfortunately, due to the cognitive complexity and the lack of involuntary cues related to lying features, the performances of these approaches suffer and their generalization ability is limited. To improve performance, this study proposed state transition patterns based on hands, body motions, and eye blinking features from real-life court trial videos. Each video frame is represented according to a computed threshold value among neighboring pixels to extract spatial–temporal state transition patterns (STSTP) of the hand and face poses as involuntary cues using fully connected convolution neural network layers optimized with the weights of ResNet-152 learning. In addition, this study computed an eye aspect ratio model to obtain eye blinking features. These features were fused together as a single multi-modal STSTP feature model. The model was built using the enhanced calculated weight of bidirectional long short-term memory. The proposed approach was evaluated by comparing its performance with current state-of-the-art methods. It was found that the proposed approach improves the performance of detecting lies.

Backhand-Approach-Based American Sign Language Words Recognition Using Spatial-Temporal Body Parts and Hand Relationship Patterns

Most of the existing methods focus mainly on the extraction of shape-based, rotation-based, and motion-based features, usually neglecting the relationship between hands and body parts, which can provide significant information to address the problem of similar sign words based on the backhand approach. Therefore, this paper proposes four feature-based models. The spatial–temporal body parts and hand relationship patterns are the main feature. The second model consists of the spatial–temporal finger joint angle patterns. The third model consists of the spatial–temporal 3D hand motion trajectory patterns. The fourth model consists of the spatial–temporal double-hand relationship patterns. Then, a two-layer bidirectional long short-term memory method is used to deal with time-independent data as a classifier. The performance of the method was evaluated and compared with the existing works using 26 ASL letters, with an accuracy and F1-score of 97.34% and 97.36%, respectively. The method was further evaluated using 40 double-hand ASL words and achieved an accuracy and F1-score of 98.52% and 98.54%, respectively. The results demonstrated that the proposed method outperformed the existing works under consideration. However, in the analysis of 72 new ASL words, including single- and double-hand words from 10 participants, the accuracy and F1-score were approximately 96.99% and 97.00%, respectively.