People with hearing impairments are found worldwide; therefore, the development of effective local level sign language recognition (SLR) tools is essential. We conducted a comprehensive review of automated sign language recognition based on machine/deep learning methods and techniques published between 2014 and 2021 and concluded that the current methods require conceptual classification to interpret all available data correctly. Thus, we turned our attention to elements that are common to almost all sign language recognition methodologies. This paper discusses their relative strengths and weaknesses, and we propose a general framework for researchers. This study also indicates that input modalities bear great significance in this field; it appears that recognition based on a combination of data sources, including vision-based and sensor-based channels, is superior to a unimodal analysis. In addition, recent advances have allowed researchers to move from simple recognition of sign language characters and words towards the capacity to translate continuous sign language communication with minimal delay. Many of the presented models are relatively effective for a range of tasks, but none currently possess the necessary generalization potential for commercial deployment. However, the pace of research is encouraging, and further progress is expected if specific difficulties are resolved.

/pdf/deep-learning-for-sign-language-recognition-current-25egznqe9z.pdf

Deep Learning for Sign Language Recognition: Current Techniques, Benchmarks, and Open Issues

Chinese Sign Language (CSL) offers the main means of communication for the hearing impaired in China. Sign Language Recognition (SLR) can shorten the distance between the hearing-impaired and healthy people and help them integrate into the society. Therefore, SLR has become the focus of sign language application research. Over the years, the continuous development of new technologies provides a source and motivation for SLR. This paper aims to cover the most recent approaches in Chinese Sign Language Recognition (CSLR). With a thorough review of superior methods from 2000 to 2019 in CSLR researches, various techniques and algorithms such as scale-invariant feature transform, histogram of oriented gradients, wavelet entropy, Hu moment invariant, Fourier descriptor, gray-level co-occurrence matrix, dynamic time warping, principal component analysis, autoencoder, hidden Markov model (HMM), support vector machine (SVM), random forest, skin color modeling method, k-NN, artificial neural network, convolutional neural network (CNN), and transfer learning are discussed in detail, which are based on several major stages, that is, data acquisition, preprocessing, feature extraction, and classification. CSLR was summarized from some aspect as follows: methods of classification and feature extraction, accuracy/performance evaluation, and sample size/datasets. The advantages and limitations of different CSLR approaches were compared. It was found that data acquisition is mainly through Kinect and camera, and the feature extraction focuses on hand’s shape and spatiotemporal factors, but ignoring facial expressions. HMM and SVM are used most in the classification. CNN is becoming more and more popular, and a deep neural network-based recognition approach will be the future trend. However, due to the complexity of the contemporary Chinese language, CSLR generally has a lower accuracy than other SLR. It is necessary to establish an appropriate dataset to conduct comparable experiments. The issue of decreasing accuracy as the dataset increases needs to resolve. Overall, our study is hoped to give a comprehensive presentation for those people who are interested in CSLR and SLR and to further contribute to the future research.

A Survey on Artificial Intelligence in Chinese Sign Language Recognition

This paper presents a novel vehicle detection approach for detecting vehicles from static images using color and edges. Different from traditional methods, which use motion features to detect vehicles, this method introduces a new color transform model to find important "vehicle color" for quickly locating possible vehicle candidates. Since vehicles have various colors under different weather and lighting conditions, seldom works were proposed for the detection of vehicles using colors. The proposed new color transform model has excellent capabilities to identify vehicle pixels from background, even though the pixels are lighted under varying illuminations. After finding possible vehicle candidates, three important features, including corners, edge maps, and coefficients of wavelet transforms, are used for constructing a cascade multichannel classifier. According to this classifier, an effective scanning can be performed to verify all possible candidates quickly. The scanning process can be quickly achieved because most background pixels are eliminated in advance by the color feature. Experimental results show that the integration of global color features and local edge features is powerful in the detection of vehicles. The average accuracy rate of vehicle detection is 94.9%

以正規化色彩與邊緣資訊作車輛偵測; Vehicle Detection Using Normalized Color and Edge Map

Sign language is the major means of communication for the deaf community. It uses body language and gestures such as hand shapes, lib patterns, and facial expressions to convey a message. Sign language is geography-specific, as it differs from one country to another. Arabic Sign language is used in all Arab countries. The availability of a comprehensive benchmarking database for ArSL is one of the challenges of the automatic recognition of Arabic Sign language. This article introduces KArSL database for ArSL, consisting of 502 signs that cover 11 chapters of ArSL dictionary. Signs in KArSL database are performed by three professional signers, and each sign is repeated 50 times by each signer. The database is recorded using state-of-art multi-modal Microsoft Kinect V2. We also propose three approaches for sign language recognition using this database. The proposed systems are Hidden Markov Models, deep learning images’ classification model applied on an image composed of shots of the video of the sign, and attention-based deep learning captioning system. Recognition accuracies of these systems indicate their suitability for such a large number of Arabic signs. The techniques are also tested on a publicly available database. KArSL database will be made freely available for interested researchers.

KArSL: Arabic Sign Language Database

In this paper, we propose a deep learning for sign language recognition using convolutional neural network (CNN) and long short term memory (LSTM). The architecture used CNN as a pretrained model for feature extraction and is passed to the LSTM for capturing spatio-temporal information. One more LSTM is stacked for increasing the accuracy. Deep learning model which captures temporal information is less. There is only less papers which deals with sign language recognition by using the deep learning architectures such as CNN and LSTM. The algorithm was tested in Indian sign language (ISL) dataset. We have presented the performance evaluation after testing with ISL dataset. Literature shows that deep learning models capturing temporal information is still an open research problem.

CNN and Stacked LSTM Model for Indian Sign Language Recognition

The Protection of information with steganography is the art and science of writing hidden is a hidden data method in various data type to conceal data and unauthorized person or attacker does not know these messages are camouflage. Moreover, we can send text or image was concealed by cover image and adjust some bits of the original to the receiver. The text or image file concealed within cover image can secure the original data from the attacker. In this paper, we studied the performance of the optimal file size to camouflage and the hidden data method with cover image using Least Significant bit (LSB). The dataset are text file (.txt), document File (.doc) and image file which have dimension as follow; 800x600, 1200x900, 1600x1200, 2000x1500, 2400x1800, 2800x2100, 3200x2400, 3600x2700 and 4000x3000 The image file sizes are 576KB, 1.14MB, 1.85MB, 2.66MB, 3.58MB, 4.57MB, 5.54MB, 6.46MB and 7.31MB respectively, that used as cover image which was concealed and decrypted from Stego image to plaintext. The result showed that the cover image size is 576 KB is the optimal file for hidden data 175 KB or 30% of the cover image. The cover image size is 4.57 MB is the optimal file for hidden data 2.10 MB or 45.95% of the cover image. And the cover image size is 7.31 MB which dimension is 4000*3000 can camouflage file up to 4.29 MB or 58.69% of the cover image. Therefore, the larger cover image can hidden larger text file up to size of cover image vary to the size of cover image that used for hidden data.

The Performance of Cover Image Steganography for Hidden Information within Image File using Least Significant bit algorithm

Traffic jams are a major problem in the lives of people in the capital. The occurrence of such problems is due to drivers who do not respect traffic laws, inappropriate behavior of the driver, and the cause of illegal parking at prohibited parking, resulting in the rear car unable to move further. Possible or the car must change lanes to another lane. Therefore, to reduce such problems, we have developed an illegal parking pattern recognition system for cars on the road. We have applied the Support Vector Machine (SVM) to the signal from CCTV cameras and find important characteristics of illegal car parked at the bus stop. Then use those images to learn to recognize the parking behavior of cars with Linear Regression. From the experiment, it is found that the recognition of cars parked on the road in front of the bus stop has an accuracy rate of 82.22 percent which can be used to detect the soaking of personal cars in real life.

Recognizing the Illegal Parking Patterns of Cars on the Road in Front of the Bus Stop Using the Support Vector Machine

Sign language is a language that uses hand gesture to present a word within the hearing impairment communication. But, the important problem is sign language is not usual for ordinary people which is misunderstand meaningful. However, the advancement in computer vision and recognition process could be adapt in computer system to learn or recognize sign language. These are useful for the system to recognized and translated sign language which ordinary people understand. In this article presents a study in process and method which related with the recognition of sign language using deep learning. The algorithm is 3D-CNN used for recognized process and the images received from Kinect sensor. We found that the sign language recognition using 3D-CNN was effective, the highest accuracy of the recognition was 91.23%.

The Efficiency of Sign Language Recognition using 3D Convolutional Neural Networks

This paper presents the performance of crop yield forecasting model by comparing the accuracy of the crop yield forecasting models based on artificial intelligence. To compare the error values, researchers use 2 statistical values which are the Coefficient of Multiple Determination (R2) and the Root Mean Square Error (RMSE). The model has the highest R2 value and the lowest RMSE value will be the appropriate model. The models used for comparison in this paper include Linear Regression, Decision Tree Regression, K-Nearest Neighbors Regression, Support Vector Regression, and Multilayer Perceptron Network. The learning dataset comes from the survey of planting area and yield of lychee in Thailand during 2004-2018 and dataset in 2019 to test the performance of the model. The results show that Support Vector Regression has the highest accuracy in forecasting, followed by Multilayer perceptron network, Linear Regression, Decision Tree Regression and k-Nearest Neighbors Regression, respectively.

The Performance of Crop Yield Forecasting Model based on Artificial Intelligence

This paper presents the performance of the super-resolution visualization by Triangular Interpolation algorithm to extract characteristic in sign language recognition. By compares performance from sign language image files in the experiment, the results showed that the generation of the super-resolution image by improved Triangulation Interpolation technique can provide the best results when evaluating image performance using PSNR which has a similarity value between the original image and the high-resolution image. These use the PSNR method for measuring image quality. The PSNR value of sign language image is 40.6081 or has more efficiency at 13.15 percent when compared with the SRCNN techniques which are closest to the original image. For measuring performance by SSIM, which is a structured similarity measurement techniques, Triangulation Interpolation method can get the results of generating the super-resolution image next below the SRCNN technique. But in case of the real-time process, Triangulation Interpolation methods can process faster.

Eakbodin Gedkhaw

Papers

The Performance of Cover Image Steganography for Hidden Information within Image File using Least Significant bit algorithm

Recognizing the Illegal Parking Patterns of Cars on the Road in Front of the Bus Stop Using the Support Vector Machine

The Efficiency of Sign Language Recognition using 3D Convolutional Neural Networks

The Performance of Crop Yield Forecasting Model based on Artificial Intelligence

A Super-Resolution Image Reconstruction using Triangulation Interpolation in Feature Extraction for automatic sign language recognition