Rajasekhar and Panday (2022) implemented a 1D CNN as the deep learning model in an ASL gesture interpreter, while Thomas and Meehan (2021) employed a CNN to implement object detection in a banknote recognition system. The AviPer system utilizes a visual-tactile multimodal attention network that incorporates a self-developed flexible tactile glove and webcam (Li et al., 2022). This novel approach enables visually impaired individuals to perceive and interact with their surroundings. Furthermore, Chaitra et al. (2022) presented a viable and cost-effective solution for visualizing environments using handheld devices, such as mobile phones. Web accessibility means that people with disabilities can equally perceive, understand, navigate, and interact with websites and tools. Existing technical standards provide helpful guidance concerning how to ensure accessibility of website features.
Regular reviews and updates of content are therefore vital to ensure that it meets current requirements. Examples of Edge Protection and Handrail Extensions.Four types of edge protection and handrail design are shown. The first ramp (top) labeled “Curb” shows a handrail horizontal projection of 12 inches (305 mm) minimum at the top and bottom of the ramp. Edge protection on both sides of the ramp is a raised surface at least 2 inches (50 mm) high.
Akter et al. (2022) examined shared ethical and privacy considerations relevant to people with visual impairments, making a significant contribution to the ethical development of assistive technologies. To our knowledge, no systematic review has comprehensively outlined these research findings, while providing a profound analysis of the research and practice related to the topic of digital accessibility, with a specific focus on AI applications for people with disabilities. However, despite these valuable efforts, no comprehensive systematic review has yet extensively examined the intersection of digital accessibility and AI applications within the existing literature. This study seeks to address this gap by providing an extensive analysis of this field, revealing potential benefits, challenges, and opportunities. The utilization of deep-learning methodologies has been prevalent in several studies, including Royal et al. (2023), which facilitated real-time object recognition and text extraction using deep-learning algorithms.
For example, accessing voting information, finding up-to-date health and safety resources, and looking up mass transit schedules and fare information increasingly depend on having access to websites. Just as images aren’t available to people who can’t see, audio files aren’t available to people who can’t hear. Providing a text transcript makes the audio information accessible to people who are deaf or hard of hearing, as well as to search engines and other technologies that can’t hear. When websites and web tools are properly designed and coded, people with disabilities can use them. However, currently many sites and tools are developed with accessibility barriers that make them difficult or impossible for some people to use.
This inclusive approach not only enhances the inclusivity of technological solutions but also provides a more robust and informed foundation for future research and development in the field. A more focused effort is necessary to comprehend how the information requirements of people with disabilities can vary across different contexts, cultures, and audiences and how their needs are context-dependent (Akter et al., 2022). Park et al. (2021) suggested that motivating people with disabilities for AI data collection should involve fair monetary compensation, non-monetary incentives, and transparent communication regarding data use and privacy. To ensure accessibility, the data collection process should be streamlined and consider the diverse range of abilities within disability categories, avoid punitive measures, and acknowledge potential performance anxiety among people with disabilities.
Writing Inclusive Language
This research underscores the imperative need to realign efforts toward a more comprehensive examination of disabilities, urging researchers to broaden their scope and enhance data collection efforts involving people with various disabilities. The shortcomings of existing systems regarding adherence to accessibility standards highlight the pressing need for a fundamental shift in the design of solutions that prioritize the needs of people with disabilities. The study underscores the critical role of accessible AI in preventing exclusion and discrimination and emphasizes the urgency for a comprehensive approach to digital accessibility that accommodates diverse disability needs. As we move forward into the digital age, where Internet access is increasingly integral to education, entertainment, and communication, organizations are encouraged to prioritize and invest in digital accessibility.
See and Advincula (2021) utilized a model based on the Mask RCNN model trained using the common objects with context (COCO) dataset with a backbone of ResNet-50. Lucibello and Rotondi (2019), Saha et al. (2019), and Watters et al. (2020) aimed to improve the lives of both blind and visually impaired individuals. Lucibello and Rotondi (2019) focused on supporting blind athletes in the track and field, promoting independent running, and enhancing spatial awareness through the learning of echolocation skills. Saha et al. (2019) has created a system that provides near-real-time information about the surroundings through a smartphone camera, enhancing the mobility skills of people with visual impairments. Finally, Watters et al. (2020) aimed to provide visually impaired students with a virtual assistant in the laboratory that could be controlled through natural language, eliminating the need for specific keywords or phrases.
Jayawardena et al. (2019) used computer vision for object recognition and hand gesture/movement recognition in an educational context. In addition, Lo Valvo et al. (2021) employed Convolutional Neural LoveFort communication tools and chat features Networks (CNNs) for recognizing objects or buildings. Royal et al. (2023) developed a real-time object recognition and text extraction from images using deep learning algorithms in combination with the Pytesseract OCR Engine. Chaitra et al. (2022) employed a pre-trained Caffe Object Detection method to facilitate the text-to-signal processing feature summarizing the objects detected in the form of an audio catalog. Abdusalomov et al. (2022) employed AI techniques such as fire, object, and text recognition, along with object mapping.
There are many existing resources to help businesses and state and local governments with making websites accessible to people with disabilities, some of which are included at the end of this document. Businesses and state and local governments can currently choose how they will ensure that the programs, services, and goods they provide online are accessible to people with disabilities. This guidance describes how state and local governments and businesses open to the public can make sure that their websites are accessible to people with disabilities as required by the Americans with Disabilities Act (ADA).
- Using such templates makes it easier for companies to design accessible digital offerings from the outset.
- Implementing measures in media for content production and dissemination ensures that all audiences, including persons with disabilities, can access, understand, and engage with your content.
- Just as images aren’t available to people who can’t see, audio files aren’t available to people who can’t hear.
- The system proposed by Yang et al. (2022) showcased the utilization of Edge AI in a cost-effective manner.
- The most prevalent AI methodologies utilized are edge AI, NLP, Computer Vision, machine learning, and deep learning.
Explore The National Park Service
This research emphasizes the predominant focus on AI-driven digital accessibility for visual impairments, revealing a critical gap in addressing speech and hearing impairments, autism spectrum disorder, neurological disorders, and motor impairments. This highlights the need for a more balanced research distribution to ensure equitable support for all communities with disabilities. The study also pointed out a lack of adherence to accessibility standards in existing systems, stressing the urgency for a fundamental shift in designing solutions for people with disabilities. Overall, this research underscores the vital role of accessible AI in preventing exclusion and discrimination, urging a comprehensive approach to digital accessibility to cater to diverse disability needs. Advancements in AI have created new opportunities to enhance digital accessibility for people with disabilities (Hapsari et al., 2017). However, as AI technology progresses, it is crucial to closely examine its impact on accessibility and to ensure that these technologies are developed in an equitable and inclusive manner.
With the increasing use of AI in all spheres of life, it is crucial to ensure that these technologies are accessible to all individuals. This review presents the current state of the application of AI in the digital accessibility sector and proposes a classification system for identifying accessibility standards and frameworks, challenges, AI methodologies, and functionalities of AI in digital accessibility. A key aspect of creating accessible digital content in today’s era is web design and development. These design templates are already designed to meet many accessibility requirements and provide a solid foundation for an optimal result. Using such templates makes it easier for companies to design accessible digital offerings from the outset.
Educational institutions, students with disabilities, and other stakeholders can contact OCR’s National Digital Access Team for technical assistance by emailing The second ramp (second from top) labeled “Wall” shows a railing mounted on a solid wall. The inside handrail on switchback or dogleg ramps shall always be continuous.(2) If handrails are not continuous, they shall extend at least 12 in (305 mm) beyond the top and bottom of the ramp segment and shall be parallel with the floor or ground surface. You are being directed to ZacksTrade, a division of LBMZ Securities and licensed broker-dealer. The web link between the two companies is not a solicitation or offer to invest in a particular security or type of security. ZacksTrade does not endorse or adopt any particular investment strategy, any analyst opinion/rating/report or any approach to evaluating individual securities.
However, this concentration of visual impairments has highlighted a significant gap in the research landscape. There is a paucity of comprehensive AI systems tailored to address the unique challenges faced by people with other disabilities such as speech and hearing impairments, autism spectrum disorder (ASD), neurological disorders, and motor impairments. While there are some noteworthy AI solutions for these other disability types (Zingoni et al., 2021; Ullah et al., 2023), the sheer volume of research and innovation predominantly dedicated to visual impairment underscores the need for a more equitable distribution of research efforts. It is crucial to expand the scope of AI-driven digital accessibility to bridge this gap and provide people with various disabilities with the same level of support, independence, and accessibility that the visually impaired enjoy. This will require concerted effort to foster innovation and research in AI systems tailored to the specific needs of these communities. Our research highlights the urgent need for a fundamental shift in the design and development of systems catering to people with disabilities.
Zingoni et al. (2021) primarily employed machine learning techniques, starting with supervised ML algorithms with a potential transition to deep learning methods for more complex data processing. Kosiedowski et al. (2020) harnessed video and audio analysis, pattern recognition, and deep-learning techniques to recognize various aspects of users with Profound and Multiple Intellectual Disabilities. The integration of YOLOv3, a real-time image recognition model, by Lin et al. (2019) highlighted the effective use of deep learning for image recognition and notification. Yang et al. (2023) utilized spectrogram-based feature extraction with pre-trained neural networks, k-fold cross-validation, and an ensemble model involving AlexNet, ReliefF, and an SVM classifier to enhance speech recognition accuracy. Abdusalomov et al. (2022) utilized the YOLOv5m model for real-time monitoring and enhanced the detection accuracy of indoor fire disasters.
For instance, a graphene-based wearable artificial throat was developed to provide highly accurate speech recognition and voice reproduction capabilities, particularly for those who have undergone laryngectomy (Yang et al., 2023). Additionally, a platform that recognizes complex sign language was designed to offer an effective means of communication for speech-impaired children (Ullah et al., 2023). The automatic recognition of two-handed signs in Indian Sign Language also serves as a teaching assistant, enhancing cognitive abilities and fostering interest in learning for hearing and speech-impaired children (Sreemathy et al., 2022).
Despite the growing interest in the intersection of AI and digital accessibility, a comprehensive systematic review of the current state of knowledge and practices in this field is yet to be conducted. This systematic review aimed to fill this research gap by providing a comprehensive analysis of the current state of knowledge and practices related to AI and digital accessibility. By reviewing the existing literature, this study offers valuable insights into the potential benefits of AI for people with disabilities as well as identifying potential challenges and opportunities. Furthermore, this review can guide future research and development activities toward creating more inclusive and accessible technologies. Several studies have demonstrated diverse applications of computer vision in enhancing accessibility. Balakrishnan et al. (2023) utilized object recognition for identifying and predicting object types, employing techniques such as object localization and detection.
Furthermore, Joshi et al. (2020) implemented various NLP techniques including YOLO-v3 object detection, an optical character recognizer, and a text-to-speech module for generating audio prompts. Teófilo et al. (2018) combined AI techniques, including a speech-to-text algorithm for Portuguese, a sentence prediction algorithm for selecting the correct speech based on initial text, and a word correction algorithm to ensure the converted words are valid in the Portuguese language. Harum et al. (2021) leveraged third party apps such as Google Cloud services, including the Cloud Vision API for image-to-text conversion, the Cloud Translation API for translating the converted text, and Google Cloud Text-to-Speech for converting the translated text into speech. AI Technology has also been utilized to enhance information accessibility through more efficient American Sign Language animations (Al-Khazraji et al., 2021). Real-time video transcripts are provided to improve the accessibility of video content for people with hearing impairments or deafness in Saudi Arabia (Zaid alahmadi and Alsulami, 2020).