Research Article
Behavior learning via state chains from motion detector sensors
@INPROCEEDINGS{10.4108/ICST.BIONETICS2007.2423, author={Dietmar Bruckner and Brian Sallans and Roland Lang}, title={Behavior learning via state chains from motion detector sensors}, proceedings={2nd International ICST Conference on Bio-Inspired Models of Network, Information, and Computing Systems}, proceedings_a={BIONETICS}, year={2008}, month={8}, keywords={Hidden Markov models behavior recognition intelligent environment intelligent sensor systems}, doi={10.4108/ICST.BIONETICS2007.2423} }- Dietmar Bruckner
Brian Sallans
Roland Lang
Year: 2008
Behavior learning via state chains from motion detector sensors
BIONETICS
ICST
DOI: 10.4108/ICST.BIONETICS2007.2423
Abstract
A method for the automatic discrimination of sensor and system behavior and construction of symbols with semantic meaning from simple sensor data is introduced - SCRS (Semantic Concept Recognition System). The automated method is based on statistical modelling of sensor behavior. A model of a sensorpsilas value sequences is automatically constructed. The modelpsilas structure and parameters are optimized using an minibatch model merging and parameter updating algorithm. Incoming sensor values are then conveyed to the model and the most probable path through the model to some particular state is computed. That classification can be interpreted as a semantic symbol or concept. The SCRS can be used as a security and care system for observation of persons and interpretation of scenarios. An example modelling a motion detector is discussed. The SCRSpsilas method of representing scenarios can be also used in autonomous agents for decision making processes. An example is discussed.
