Search

A review of modern methods and technologies for automatic machine translation for the deaf and hard of hearing is presented, including recognition and synthesis of both spoken and sign languages. These methods aim to facilitate effective communication between deaf/hard-of-hearing and hearing individuals. The proposed solutions have potential applications in contemporary human-machine interaction interfaces. Key aspects of new technologies are examined, including methods for sign language recognition and synthesis, audiovisual speech recognition and synthesis, existing corpora for training neural network models, and current systems for automatic machine translation. Current neural network approaches are presented, including the use of deep learning methods such as convolutional and recurrent neural networks as well as transformers. An analysis of existing corpora for training recognition and synthesis systems is provided, along with an evaluation of the challenges and limitations of existing machine translation systems. The main shortcomings and specific problems of current automatic machine translation technologies are identified, and promising solutions are proposed. Special attention is given to the applicability of automatic machine translation systems in real-world scenarios. The need for further research in data collection and annotation, development of new methods and neural network models, and creation of innovative technologies for processing audio and video data to enhance the quality and efficiency of the existing automatic machine translation systems is highlighted.

Visual speech recognition or automated lip-reading systems actively apply to speech-to-text translation. Video data proves to be useful in multimodal speech recognition systems, particularly when using acoustic data is difficult or not available at all. The main purpose of this study is to improve driver command recognition by analyzing visual information to reduce touch interaction with various vehicle systems (multimedia and navigation systems, phone calls, etc.) while driving. We propose a method of automated lip-reading the driver’s speech while driving based on a deep neural network of 3DResNet18 architecture. Using neural network architecture with bi-directional LSTM model and attention mechanism allows achieving higher recognition accuracy with a slight decrease in performance. Two different variants of neural network architectures for visual speech recognition are proposed and investigated. When using the first neural network architecture, the result of voice recognition of the driver was 77.68 %, which was lower by 5.78 % than when using the second one the accuracy of which was 83.46 %. Performance of the system which is determined by a real-time indicator RTF in the case of the first neural network architecture is equal to 0.076, and the second — RTF is 0.183 which is more than two times higher. The proposed method was tested on the data of multimodal corpus RUSAVIC recorded in the car. Results of the study can be used in systems of audio-visual speech recognition which is recommended in high noise conditions, for example, when driving a vehicle. In addition, the analysis performed allows us to choose the optimal neural network model of visual speech recognition for subsequent incorporation into the assistive system based on a mobile device.

This paper aims to investigate the possibility of robustness enhancement as applied to an automatic system for isolated signs and sign languages recognition, through the use of the most informative spatiotemporal visual features. The authors present a method for the automatic recognition of gestural information, based on an integrated neural network model, which analyses spatiotemporal visual features: 2D and 3D distances between the palm and the face; the area of the hand and the face intersection; hand configuration; the gender and the age of signers. A 3DResNet-18-based neural network model for hand configuration data extraction was elaborated. Deepface software platform neural network models were embedded in the method in order to extract gender and age-related data. The proposed method was tested on the data from the multimodal corpus of sign language elements TheRuSLan, with the accuracy of 91.14 %. The results of this investigation not only improve the accuracy and robustness of machine sign language translation, but also enhance the naturalness of human-machine interaction in general. Besides that, the results have application in various fields of social services, medicine, education and robotics, as well as different public service centers.