Using Sensor Sequences for Activity Recognition by Mining and Multi-Class Adaboost

- In this paper, we present an activity recognition system using sensor sequence information generated from many binary on-off state sensors. When many sensors are deployed the number of sensor activation sequence or any combination will be huge and handling the combination is beyond the capability of traditional classifiers. We propose to mine important features from training data set and use multiclass Adaboost to further reduce the dimension. As a result we get a real-time lightweight classifier for activity recognition.

• Type: Journal
• Year: 2010
• Where: ICAI
• Authors: Md. Kamrul Hasan, Sungyoung Lee, Young-Koo Lee
• Keywords: Activity Recognition, Artificial Intelligence, ICAI 2010, Sensor, Sensor Activation Sequence