Lightweight Sensor Data Fusion Based on Dynamic Correlation Maintence - A Case Study on Smart Home

Handuo Zhang, Weiye Xie, Boyang Yin, Jun Na, Bin Zhang

Abstract: In the Internet of Things(IoT) environment, the exponential growth of variety and number of sensors brings computing and transmission pressure to data fusion. In general, reducing some dimensions and using related dimension substitutes can significantly reduce the amount of data collected and transmitted, thus improving the fusion efficiency. Traditional methods typically capture a static correlation relationship and model complex substitution function between sensors. However, the correlation relationship commonly changes, as demonstrated in our smart home case study. Therefore, a lightweight yet dynamic correlation modeling approach is more efficient. Aiming to capture and maintain dynamic correlation for data fusion, a Lightweight Sensor Data Fusion based on Dynamic Correlation Maintence Algorithm is proposed. It comprises two stages: a static configuration stage and a dynamic reduction stage It comprises two stages: a static configuration stage and a dynamic reduction stage. In the static configuration stage, sensors are grouped based on their data correlation, and substitution functions are generated by linear regression. In the dynamic reduction phase, changes in sensor correlation are detected using concept drift detection, triggering the regeneration of grouping or substitution functions. Case study on smart home and experimental analyses show that our algorithm achieves higher data reduction rates and accuracy than static correlation and temporal correlation methods.