Multi-agent Negotiation Mechanisms for Statistical Target Classification in Wireless Multimedia Sensor Networks

Author： Wang Xue Bi Dao-wei Ding Liang Wang Sheng

Publisher： MDPI

E-ISSN： 1424-8220|7|10|2201-2237

ISSN： 1424-8220

Source： Sensors, Vol.7, Iss.10, 2007-10, pp. : 2201-2237

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

The recent availability of low cost and miniaturized hardware has allowed wireless sensor networks (WSNs) to retrieve audio and video data in real world applications, which has fostered the development of wireless multimedia sensor networks (WMSNs). Resource constraints and challenging multimedia data volume make development of efficient algorithms to perform in-network processing of multimedia contents imperative. This paper proposes solving problems in the domain of WMSNs from the perspective of multi-agent systems. The multi-agent framework enables flexible network configuration and efficient collaborative in-network processing. The focus is placed on target classification in WMSNs where audio information is retrieved by microphones. To deal with the uncertainties related to audio information retrieval, the statistical approaches of power spectral density estimates, principal component analysis and Gaussian process classification are employed. A multi-agent negotiation mechanism is specially developed to efficiently utilize limited resources and simultaneously enhance classification accuracy and reliability. The negotiation is composed of two phases, where an auction based approach is first exploited to allocate the classification task among the agents and then individual agent decisions are combined by the committee decision mechanism. Simulation experiments with real world data are conducted and the results show that the proposed statistical approaches and negotiation mechanism not only reduce memory and computation requirements in WMSNs but also significantly enhance classification accuracy and reliability.