EUR
en
- es
- pt
- fr
- de
- it
Published on 2024-01-15
Personal assistants like Siri or Alexa allow turning on the TV or turning off the light just by asking. However, they have trouble understanding a person with speech difficulties. The same goes for applications that allow scanning text but do not read it aloud, which is indispensable for someone with visual impairment. Artificial intelligence can be a support to overcome these barriers, like programs trained to process all kinds of speeches and voice inflections or smart glasses that can describe our surroundings.
Some uses of artificial intelligence in recent years focus on generating images and videos from text instructions, known as 'prompts.' But there are applications that enable the reverse process: uploading a photograph or recording a video for artificial intelligence to describe the elements appearing in the image.
This is the case with Be My Eyes, an application that since August 2023 has incorporated this technology to describe images for visually impaired users. Until now, it was necessary to wait to contact a volunteer through this platform to receive this type of assistance, but with artificial intelligence, it is simpler.
The GPT-4 language model, the same technology used by ChatGPT, allows requesting this support from an artificial intelligence and asking for help in finding dropped objects, reading labels, or guiding us through a place. Be My Eyes is available in the Play Store and the App Store.
Ask Envision is another project using ChatGPT technology to provide new tools for people with vision problems. Instead of a mobile app, Ask Envision uses Envision glasses: internet-connected glasses based on Google Glass technology that allows taking images of our surroundings and giving us a description of the elements around us.
For example, it is possible to scan a restaurant menu and have it read the entire menu or ask the assistant what vegetarian options are available. The Envision glasses are available from 1,899 euros in their basic version.
Artificial intelligence has its place in the world of sound and speech. Just like applications such as Google Translate, which allows translating any conversation in real time (a useful tool for asking for directions in other languages if we travel), Google Live Transcript (available in the Play Store) performs a real-time transcription of the conversations around us and alerts us to sounds like a dog barking or a doorbell ringing.
These translation tools can also be used in sign languages. This is the project of Priyanjali Gupta, an engineering student at the Vellore Institute of Technology in India, who has developed an artificial intelligence capable of translating American Sign Language (ASL, used in the United States, Canada, and Mexico) into written English in real time.
Artificial intelligence can be used as support to develop different competencies or as a therapeutic resource for children with speech difficulties, like Timlogo. It is an online platform (available only in Romanian) developed by the Romanian company Ascendia that offers games for underage patients and is capable of analyzing children's pronunciation to detect specific problems in their speech and offer personalized exercises through this technology.
People with hearing impairment may have speech difficulties, which can hinder oral communication with others. Although there are tools capable of converting written text into speech (a process known as 'text-to-speech'), these systems are not entirely fluent in maintaining a conversation.
To try to solve this problem, Google has developed Parrotron: an artificial intelligence capable of recognizing the speech of a person with speech difficulties and making a synthetic voice (generated by computer and capable of imitating our way of speaking) repeat each phrase with a consistent cadence and prosody, although it is currently in the research phase.
Speech problems can make it difficult to use other technology like voice assistants, such as Siri, Alexa, or Google Assistant, which can be useful for people with certain types of disabilities as they allow us to call our contacts, listen to a podcast, or turn on the lights in a room with just a voice command.
Although these systems are trained with hundreds of voice recordings to learn to recognize oral instructions, they are not prepared to understand the speech of someone with speech difficulties. But gradually alternatives are emerging like Voiceitt: an artificial intelligence model focused on users with speech difficulties that can be implemented in voice assistants like Alexa or used to generate transcriptions in online meetings on platforms like Webex. Voiceitt is available in Spanish through a monthly subscription of 50 dollars on its web platform.
There are cases where a person may have permanently lost the ability to speak, such as patients with Amyotrophic Lateral Sclerosis (ALS). Although text-to-speech conversion systems can facilitate communication for those who no longer retain speech, these tools use synthetic voices that can seem impersonal and artificial.
Thanks to artificial intelligence and cloned voices (those that mimic the speech, timbre, and prosody of a particular person), personalized text-to-speech conversion systems can be created that help patients maintain their identity and improve their quality of life. This was the case with the father of Álvaro Medina, a journalist at Prodigioso Volcán, who in collaboration with ahoLab managed to preserve his voice after recording different samples of his voice with a microphone and creating an acoustic model of it, a function now available on Apple devices.
COMMENTS
No customer comments for the moment.