FONG Kasin is a new media technologist who mainly interested in computational installation,
interaction, generative coding, physical computing and machine learning. With a background
in computer science, Kasin previously worked as an interactive developer at digital agencies
in Hong Kong. Her works are diverse and include interactive installations, projection mapping,
games and animation.
More on her personal website
Emotion Blower is an interactive sound installation made up of colorful paper cups and
balloons. Working together with the machine learning application, visitors speak to the
microphone, the Emotion Blower will then select the emotion with the highest classification
rate and blow the corresponding balloon. An emotion category is represented by a distinct
color.
This project is first inspired by the song ‘First Time in Love’ from a Korean musical ‘Maybe
Happy Ending’. It is amazing that emotion, including those tiny changes, of the character can
be felt thoroughly without knowing the lyrics or seeing facial expression. I then start thinking
of keywords like emotion, sound and human voice.
Generally, we use diverse pitch, pace and tone when we are talking to different people. Voice
does not only deliver messages, but also conceal the mood and reveal the psychological
distances between the speaker and the listener. Isn’t machine learning the most direct way to
figure out the emotions hidden in human voice?
The artwork uses EmoVoice, a real-time emotions recognition framework based on acoustic
properties of speech, and trained using the RAVDESS dataset. RAVDESS contains 7356 audio
files with 8 labelled emotions (neutral, calm, happy, sad, angry, fearful, disgust, surprised),
vocalizing two statements in North American accent.