bSpace: A Data Cleaning Approach for RFID Data Streams Based on Virtual Spatial Granularity

Abstract

RFID holds the promise of real-time identifying, locating, tracking and monitoring physical objects without line of sight, and can be used for a wide range of pervasive computing applications. To achieve these goals, RFID data have to be collected, filtered, and transformed into semantic application data. RFID data, however, contain false readings and duplicates. Such data cannot be used directly by applications unless they are filtered and cleaned. To compensate for the inherent unreliability of RFID data streams, most RFID middleware systems employ a “smoothing filtering”. In this paper, a new “smoothing filtering” approach named bSpace is proposed, which is based on the concept of virtual spatial granularity. For providing accurate RFID data to applications, bSpace uses a Bayesian estimation algorithm to fill up false negatives, and uses the rules which we define to solve false positives.