PECO: a Personalized activity ECOsystem

Unmet Need: Activity Recognition in multi-view ecosystems

Smart Environments have become more ubiquitous in recent years and with these advancements, the activity recognition algorithms have advanced, but still lack the ability to collaborate and transfer information across heterogenous devices.

The Technology: Collegial Activity Learning between Heterogeneous Sensors

WSU Researchers have developed a novel system for learning and transferring the data from different sensors to improve the way in which the sensors interact within the smart environment. This method employs machine learning techniques on new but similar tasks that are collected via these environmental sensors. This technique, known as transfer learning, helps automate the processing and synthesizing of these data streams to create a systemic performance improvement. This Personal ECOsystem, or PECO, allows a user to use a smart phone to leverage the preexisting sensors within a smart environment to act as a teacher for the new smart phone and for the phone to in turn boost the performance of the smart home-based model.

Applications:

• Personal Smart Homes
• Warehouse Operations 
• Manufacturing Factories

Advantages:

• PECO provides for continually improved activity recognition using substantially less human labeled data.
• Significantly lowers the barrier to deploying and updating activity recognition systems with new sets of sensors and inputs.

Patent Information:

Patent App. No.: 14/720,078