VoiceInterruptionRecognizer - Invent a voice algorithm to recognize external interruptions

A significant part of a developer’s work flow is disturbed by interruptions from co-workers or phone calls. While these are important for the overall team productivity and team work, some of them are decreasing a developer’s productivity if they occur in an inopportune moment. In our work, we are creating innovative approaches to avoid/streamline certain interruptions. To improve the success of our tools, we need to accurately recognize external interruptions, which could be the topic of your next master project/master thesis.

Your job would be to implement and evaluate a system which uses modern voice recognition technologies to identify and analyze interruptions from co-workers and analyze a developer’s social environment, i.e. who is talking to whom and for how long, etc. It could further also help to identify times where the background noise is too high to focus on work, and hence is decreasing productivity. To achieve this, you could use a commercial microphone or build your own and implement an algorithm to distinguish different people's voices from each other.

The main tasks of this projects are:

• Use a commercial microphone or build your own (to be discussed) to capture voices.
• Use voice recognition software and train them to identify one or more persons, to know who is talking to whom and for how long.
• Extend your algorithm to understand when a discussion between people is an interruption and/or to recognize background noise.
• Provide the data and algorithm as an easy to use library, which could be used in a future study.
• Evaluate your prototype using acceptance tests for multiple people in the same room.
• Bonus: Visualize who is talking to whom by showing a different color per person on a simple light (e.g. the BlyncLight).

Posted: 04.04.2017

Contact: André Meyer and Manuela Züger