A mobile OS that helps the hearing-impaired speak more clearly
(Oct. 2024) Isn’t it frustrating when you know exactly what you want to do, but for some reason or other, you just can’t get it right? What if you had no direct way to assess the results of your efforts, and instead had to rely on others people’s reaction?
This is the constant challenge of people who suffer from hearing loss. Even if they learn how to speak with a speech therapist, it is hard or even impossible for them to judge if their speech is landing correctly. The look in the eyes of the person they’re addressing is their first clue to whether they’ve been understood.
The short video “Find your Voice” illustrates the relevance of the solution to the speech-impaired
Enter Celia Voice Enhancement, a feature of Huawei’s mobile OS HarmonyOS NEXT. Celia Voice Enhancement “translates” into a standard form the imperfect speech that is common among the hearing impaired. And not only that, rather that generating a disembodied mechanical voice, Celia emulates the unique timbre and character of the hearing-impaired user, empowering him or her to “find their voice” and correctly express a key trait of their unique personality.
When a speech-impaired person first uses Celia Voice Enhancement, all he or she needs to do is to say a few sample sentences that enable the AI to individualize the solution. From then on, the user simply speaks into his or her smartphone when addressing someone, and Celia retells the words into a standard form that anyone can understand.
A user providing a voice sample to Celia
Celia Voice Enhancement is based on a large voice model timbre cloning technology that Huawei developed for people with speech impairments. Dozens of Huawei R&D staff put their heart into this project for over one year, collecting training data, developing the engine, and implementing the project from start to finish. The project started testing with hearing impaired volunteers in 2023.
One particular challenge during development was the relatively poor availability of Chinese speech-impaired voice samples. However, the team managed to piece together over 10,000 hours of training data, enabling the technology to reach a high level of accuracy.
Speech-impaired parents don’t need to fear teaching incorrect speech to their children
In China, where HarmonyOS NEXT Celia Voice Enhancement is initially made available, about 7 million people are speech-impaired and 27 million are hearing-impaired. Globally, about 450 million people, representing 5% of the world population, suffer from hearing loss. By 2050, the World Health Organization predicts that 700 million will be affected.
Currently, Huawei’s R&D team is still working on it Celia Voice Enhancement to further raise accuracy and deliver a smoother experience to those in need.
