August 18, 2019

The sounds of routine household noises is something many of us take for granted. For people who are deaf and hard of hearing, this can be challenging. A startup called Wavio based in the United States promises to change with a device that will help deaf people see sounds.

The beep of a microwave, a baby crying, or the sound of water overflowing are noises that hearing people take for granted. That’s not the case for people who are deaf and hard of hearing, and this can be sometimes even be life threatening. An app called Wavio, developed by a startup of the same name aims to change that by helping deaf people “see” sounds”.

Developed by three deaf men, Wavio uses artificial intelligence to identify over 500 sounds.

How Wavio works

Wavio works on its own as an app or in union with a smart home device called See Sound.

  • When a sound occurs inside or close to a household, the nearest See Sound registers the increase in volume and lights up.
  • See Sound then interprets the sound it hears and makes a prediction.
  • It then visually alerts the user that the sound has occurred on their smart devices via Wi-Fi.

The system listens for sounds like a dog’s bark, baby’s cry, fire alarm alerts and even the shattering of glass.

The current products in the market are extremely limited. We can buy a device for a smoke detector and another that tells us when the doorbell is ringing but nothing that can distinguish between a microwave beeping, baby crying or a dog barking. The reason why no one has been able to create a product like this is shortage of data. In order to teach a machine learning model how to distinguish sounds with any level of accuracy you would need millions of sound samples. –Spencer Montan, Co-founder, Wavio

The answer, as the founders discovered, lay in YouTube which contains a billion hours of videos with sounds. “We trained the See Sound machine with sound clips from YouTube”, says Montan. “Over two million videos were analysed, categorised and converted into 10 second sound clips, each with a distinct sound”.

Every type of sound in the data model like yelling is made of several thousand YouTube audio samples ranging from common household noises to alerts that could mean life or death. “Every sound we add makes our machine protocol learning even smarter allowing for even higher accuracy range”, adds Montan.

At any given time, See Sound is always listening for the over 75 sounds it is programmed to hear. Once the sound occurs it interprets it and communicates it to the user.

Wavio is still being tested and if all goes well, will be made available to users by early 2020, offering nearly nine million deaf households in the United States the possibility of seeing sound.

Chennai deaf teacher Vidya Menon says a device like Wavio could change lives of deaf and hard of hearing people in India too. “I had this experience using a pressure cooker once when I left it unattended for a few seconds and when I came back, the whistle broke out due to less water and all the stream came right up to the ceiling”.

Vidya says she now never leaves the kitchen for long if there’s something cooking. “I watch over everything that’s cooking, whether its in the pressure cooker, microwave, even the oven. I cannot leave the kitchen unattended when I am alone. I have to be back every few seconds”.

A system like See Sound, says Vidya, would enable greater independence. “Now when I hear any sound through the hearing aid, I check the door first to see if someone’s there because I can’t recognise the sound type. I will be happy if devices like See Sound are launched here as it will enable the deaf community to live easily”.