Humanoid Robot Learns Object Categories using Audio and Video

A humanoid robot interacts with objects to learn object categories. The robot captures audio and video as it performs six different exploratory behaviors. A separate object categorization is formed for each behavior and sensory modality combination. The resulting 12 categorizations are unified into a single one.

This research is motivated by how infants learn some object categories. Infants learn from the events that capture their attention, like sounds and unexpected movement patterns. As they gain more experience observing these types of events with certain objects, they begin to form object categories.

This research was publish as Chapter 5 of my M.S. Thesis. It was also published in the 2011 IEEE Transactions on Autonomous Mental Development, the 2010 IEEE International Conference on Robotics and Automation, and the 2010 National Conference on Artificial Intelligence.

Category:

Science & Technology

Tags:

• Robot
• artificial intelligence
• developmental robotics
• machine learning
• computational perception
• developmental psychology

[Johnylogic16]

This is cool.