Sarthak: So when we got this project we had to
think about the most important problem that could be solved using artificial
intelligence. What we came up with is depression. Depression is a serious
medical illness that negatively affects the way you think, feel and act.
Depression has affected over 3.1 million students, in the United States alone. Vivek: Artificial Intelligence makes it
possible for machines to learn from experience, adjust to new inputs and
perform human-like tasks. Vivek: So we wanted to make an AI solution that
can help both psychiatrists and mental health professionals do their jobs much
more better. Sarthak: So what we did is we built an AI that
can listen to your voice and then detect whether you have depression or not.
Sarthak: So now I’ll be speaking to the AI and it will detect whether I have depression or not.
I am very sad Vivek: The depression AI consists of two
different classifiers. One for audio and one for text. zthe audio classifier
analyzes how your voice sounds like. It is trained on a data set where different
voice artists speak with different emotions.
It works by converting the audio into Mel’s frequency cepstrum that gives
important information on how the voice sounds like. This information is then fed
into a classifier for training. The text classifier analyzes how you speak,
irrespective of how you sound like. It is trained on a dataset where text is
labeled with its relevant emotion. The classifier then goes ahead to
pre-process the text. This is done by removing stop words such as “a”, “in” and “the” that contribute no emotion to the text; and lemmatizing meaning words conveying
the same emotion are mapped into one word. This provides adequate features
that can be used to predict the sentiment of the text. This can then be
fed into the classifier for training. And that’s it! Sarthak: Now we understand that text
and speech are not the only parameters by which depression can be detected.
Behavior, feelings and emotion- all come into play. Unfortunately our AI cannot
factor these in at this moment. Vivek: Technology has evolved
– not only recording speech but also blood rate, pulse rate, heart rate and so on. If
we were to implement this AI on a large scale we can expect much more better and convincing results.