Sensor event flow segmentation method based on behavior characteristic probability distribution constraint

Abstract

The invention provides a sensor event flow segmentation method based on behavior characteristic probability distribution constraint. The method comprises the steps of calculating behavior probabilitycharacteristics; calculating transition probability distribution of any two behaviors in the training set by taking the day as a unit; calculating probability distribution of occurrence times of eachbehavior every day; calculating the probability distribution of the number of each behavior trigger sensor; calculating the probability distribution of the type and quantity of each behavior trigger sensor; calculating probability distribution of the number of behavior instances of each cluster; calculating probability distribution of the number of behavior instances of each cluster; a set of probabilistic features of a behavior is formed and a sensor event stream is segmented. According to the resident daily behavior identification method based on the behavior characteristic probability distribution constraint, the probability distribution of multiple behavior characteristics is calculated, and the behavior instance segmentation result of the optimal sensor event meeting the characteristic set probability distribution can be obtained by applying a probability constraint solving method.

Description

Technical field [0001] The present invention relates to the technical field of sensor event segmentation, and in particular, to a sensor event stream segmentation method based on behavior feature probability distribution constraints. Background technique [0002] The daily behavior of residents will continuously trigger a group of sensors. Ideally, the stream of continuously triggered sensor events should be segmented according to complete behavior instances. However, the main method of current segmentation of sensor event streams is to first specify a time window length or spatial window (number of sensors) length, and then divide the sensor event stream equally according to the window length. The original feature of daily behavior recognition is the sensor event stream. The sensor event stream division method based on a fixed window length will cause the same behavior instance to be divided into multiple different windows, or multiple behavior instances are divided into the sam...

AI Technical Summary

Problems solved by technology

The original feature of daily behavior recognition is the flow of sensor events. The division method of sensor event flow based on fixed window length will cause the same behavior instance to be divided into multiple different windows, or multiple behavior instances to be divided into the same window. The division results of daily behaviors either lead to incomplete characteristics of daily behaviors, or are interfered by other behavioral characteristics, resulting in poor recognition of daily behaviors

Images