VoiceSpace & Visulization

VoiceSpace is a project studying visulization of remote voice conversations, and how it can be used as "an artifact of the conversation". It was mainly conducted by Prof Karrie Karahalios's past student Pooja Mathur as her masters thesis.

Three methods of visualization are used in VoiceSpace. The first one uses a speech recognition program to grab the content of the conversation and displays the words. The second one visualizes the volumn change along the time. And the third one maps one's pitch to one's volume.

In some ways, what we want to do with audio visualizing is to add more context to the spoken word such that we can extract social cues such as volume, pitch, prosody and even perhaps sentiment from it.

My Part of the Project

People's pitch changes frequently as they talk. What kind of social cue can we abstract from the pitch change in a conversation?

There is a hypothesis saying that when two people are having a conversation, if they agree with each other’s opinions, the pitches of their voice will tent to converge to each other.

My task is to design and implement an application that can visualize people’s pitch captured from a Skype conversation in real-time. We intend to use this application to verify the aforementioned hypothesis in remate audio conversations.