Tuesday, March 16, 2010

Project Status Update

This is my project description and update:

A lot of people maintain a daily journal. Although emotion can be inferred from the journal text, it still remains indirect.  Moreover, the trend in emotional change over a period of time remains hidden in the words and cannot be recognized.
 We don't really  keep track of our emotions directly. We keep track of events, problems, people, desires, dissapointments, etc wherein emotion is implicit.
For example, if one decides to be happy for a month, there is no objective way to test it.  Compounding the problem is the fact that we are ourself deluded about our own emotional state at times.
This tool will help by directly monitoring the emotional state in the speech while the user records his/her journal.

This is the workflow of my project










And every week, month a trend report will be generated highlighting the shift in emotion for the period concerned. A weekly report will look something like this:











This is work in progress. I am able to record voice, and convert speech to text after adaptation of speech grammar for a particular user. There is a shortage of good quality speaker independent speech recognition software in open source community, therefore my software will not be speaker independent. And there would be a training time for each user to properly convert speech to text.
The part of speech emotion recognition is still incomplete because I am waiting for my other class to cover some of the tools I will need for it!
I am using C++ on linux to code the software and GNUPlot to plot the trend graphs. Luscher test is going to be an optional feature for the project unless I get the book with color plates to implement it.

1 comment:

Irvika said...

Interesting. Are you using a device similar to the one Garett is creating or simply monitoring emotions based on voice recognition?