Be my ears is a free app that helps deaf users to distinguish
specific noises. When the users turn their microphone on, the app
would try to identify any dangerous sound such as a fire alarm, a
dog bark, a car’s honks etc.
Credits: Canva
Our technology
The first step to build such an app is to get pre-trained models
that can detect whether a certain noise was made. These models
generally use the ml5.soundClassifier()which allows you to
classify audio basing an estimation on sounds that have been
stored. You can learn more about this on the
ml5-website.
The model used for the app was trained with
Google's Teachable Machine. Several noises were recorded or uploaded such as the sounds
of ambulance sirens, fire alarms, dog barking and such.
Once the model is created and works properly, Google Teachable Machine
allows us to upload it in two different ways. It can be uploaded as a
drive document or as a code on P5js. The model and its data can be
found
on this drive. And its code -without the data- can be found
here.
Prospective competitors and future of the field
AI can have a sizable impact in this field and similar projects are
emerging such as
Braci's.
Even if these apps are useful in case of danger, one needs to know
that great progress has been made in the field of AI and now machine
learning is used directly on hearing aids to enhance sensorial
experience, perceive movements, orient microphones when one is talking
and such. You can consult
Vivason's website
to get more information.