Symbolic Dynamic Filtering for Pattern Recognition in Distributed Sensor Networks

This chapter focuses on feature extraction for pattern classification in distributed dynamical systems, possibly served by a sensor network. These features are extracted as statistical patterns using symbolic modeling of the wavelet images, generated from sensor time series. The proposed symbolic dynamic filtering-based feature extraction and pattern classification methodology is executable in real time on commercially available computational platforms. A distinct advantage of this method is that the low-dimensional feature vectors, generated from sensor time series in real time, can be communicated as short packets over a limited-bandwidth wireless sensor network with limited memory nodes. The sensor time-series data sets are divided into three groups: partition data, training data, and testing data. Passive infrared (PIR) sensors are widely used for motion detection. In most applications, the signals from PIR sensors are used as discrete variables. Partitioning of time series is a crucial step for symbolic representation of sensor signals.