情報処理学会 第80回全国大会講演要旨
6S-06
Preliminary investigation of co-occurrence rule extraction from multi-modal animal locomotion data
This paper performs a preliminary investigation of a method for extracting frequent patterns from multi-modal time-series data obtained from a moving object.
Due to the recent advance in sensing technologies, we can measure trajectories of animals from GPS modules as well as time-series sensor data related to locomotion from environmental and bio-sensors using small sensor nodes attached to the animals.
The proposed method first segments time-series data, i.e., trajectory and sensor data, into multiple modes, and then extracts cross-modal frequent patterns from the segmented time-series.
For example, we extract a rule describing that a ``local search'' mode extracted from trajectory data usually co-occurs with a ``stress'' mode extracted from bio-sensor data.
Due to the recent advance in sensing technologies, we can measure trajectories of animals from GPS modules as well as time-series sensor data related to locomotion from environmental and bio-sensors using small sensor nodes attached to the animals.
The proposed method first segments time-series data, i.e., trajectory and sensor data, into multiple modes, and then extracts cross-modal frequent patterns from the segmented time-series.
For example, we extract a rule describing that a ``local search'' mode extracted from trajectory data usually co-occurs with a ``stress'' mode extracted from bio-sensor data.
