Alexa, Google Assistant, Cortana: voice assistants are changing the way we shop, search, communicate or even live. At least for most people. But what about those without a voice? What about those who cannot hear? Around 466 million people worldwide have disabling hearing loss. With the SIGNS Project, we are creating awareness for digital accessibility and inclusion.
SIGNS is the first smart voice assistant solution for people with hearing loss worldwide. It’s an innovative smart tool that recognizes and translates sign language in real-time and then communicates directly with a selected voice assistant service (e.g. Amazon Alexa, Google Assistant or Microsoft Cortana). SIGNS is reinventing voice – one gesture at a time. Many people with hearing loss use their hands to speak. And that’s all they need to talk to SIGNS. How's the weather tomorrow? Change lights to blue. Find an Italian restaurant. Just speak, and SIGNS will answer.
Users can browse our most exciting stories and find out about our best achievements and internal projects, using a drag navigation principle
There are over 2 billion voice-enabled devices across the globe. Voice assistants are changing the way we shop, search, communicate or even live. At least for most people. But what about those without a voice? What about those who cannot hear? According to the World Health Organization around 466 million people worldwide have disabling hearing loss. Project SIGNS was developed to create awareness for inclusion in the digital age as well as to facilitate access to new technologies.
Many people with hearing loss use their hands to speak. Their hands are their voice. However, voice assistants use natural language processing to decipher and react only to audible commands. No sound, no reaction. Voice is predominantly seen as something audible, something you can hear. But for people with hearing loss that use their hands. Voice is more. Voice is gestures, movements, and even facial expressions. SIGNS bridges the gap between deaf people and voice assistants, by recognizing gestures to communicate directly with existing voice assistant services (e.g. Amazon Alexa, Google Home or Microsoft Cortana). SIGNS is the first smart voice assistant solution for people with hearing loss worldwide.
SIGNS was pre-trained with video footage of people whouse sign language. SIGNS includes a training interface that can be used toteach new gestures in real time. SIGNS then recognizes these gestures and acts as aninterface to voice assistant systems such as Amazon Alexa, Google Home orMicrosoft Cortana. SIGNSis based on an intelligent machine learning framework that is trained toidentify body gestures with the help of an integrated camera. These gesturesare converted into a data format that the voice assistant service understands. Thevoice assistant processes the data in real-time and replies appropriately.SIGNS replaces voice assistants’ typical form of communication through audiosignals with a visuality – but not only by displaying the word. The visualinterface of SIGNS fulfills various requirements that are necessary for anintuitive experience. SIGNS follows the basic principles of sign language.Therefore, the SIGNS dictionary was developed – a set of symbols that areinspired by the hand movements. Just like with other voice assistant devicesthe user has to naturally interact with the device.
SIGNS uses an integrated camera to recognize sign language in real-time and communicates directly with a voice assistant. The system is based on the machine learning framework Google Tensorflow. The result of the pre-trained MobileNet is used to train several KNN classifiers on gestures. The recognition calculates the likelihood of the webcam's recorded gestures and converts into text. The resulting sentences are translated into conventional grammar and sent to a cloud-based service that generates language from it. In other words, the gestures are converted into a data format (text to speech) that the selected voice assistant understands. In this case, shown Amazon Voice Service (AVS). AVS responds with meta and audio data, which in turn is converted from a cloud service to text (text to speech). The result is displayed. SIGNS works on any browser-based operating system that has an integrated camera and can be connected to a voice assistant.