Energy- Aware Signals Classification in Ad- hocWireless Sensor Networks

With the advancement of wireless and electronic technologies, wireless networks consist of tiny sensor devices hold the promise of revolutionizing sensing in a wide range of application domains because of their flexibility, low costs and ease of deployment. In this paper, the employment of ad-hoc wireless sensor networks to perform signals classification is proposed. For such application, the use of low-performance, low-power wireless sensor nodes requires the development of ad-hoc solutions of detection, features extraction and classification of the signals considered. In particular, these solutions allow to reduce the amount of data transmitted from the nodes, thus saving the consumption of energy, and the implementation costs of the classification process. Among other pattern recognition techniques based on theorems from statistical learning theory (SLT), the support vector machine is chosen for its flexibility in classifying patterns. In particular, the properties of the u-SVM allow implementing the SVM classifier on tiny sensor nodes, without significantly to make worse classification performances. As a case of study, acoustic signals are considered for implementation of the proposed algorithms on the Mical sensor node, by Crossbow Technology Inc