Transfer Learning for Activity Recognition
A key challenge that must be addressed by smart environments is the recognization and classification of user actions and activities. Currently, this occurs by training the system to recognizes the patterns of behavior of the user as they perform their activities. However, this process requires the user and/or an outside expert to initially label the activities that are being performed until the machine has "learned' the correct labelings. For each new situation, and setting this process must be repeated from scratch. The goal of transfer learning is to enable the machine to use the previously learned knowledge to correctly interpret newly encountered situations.