Authors:
Afnan Algobail
;
Adel Soudani
and
Saad Alahmadi
Affiliation:
College of Computer and Information Science and King Saud University, Saudi Arabia
Keyword(s):
WASN, Object Recognition, Acoustic Sensing, Feature Extraction, Low-power Recognition.
Related
Ontology
Subjects/Areas/Topics:
Aggregation, Classification and Tracking
;
Applications and Uses
;
Data Manipulation
;
Energy Efficiency
;
Energy Efficiency and Green Manufacturing
;
Environment Monitoring
;
Industrial Engineering
;
Informatics in Control, Automation and Robotics
;
Obstacles
;
Sensor Networks
Abstract:
Wireless Acoustic Sensor Networks (WASN) have drawn tremendous attention due to their promising potential audio-rich applications such as battlefield surveillance, environment monitoring, and ambient intelligence. In this context, designing an approach for target recognition using sensed audio data represents a very attractive solution that offers a wide range of deployment opportunities. However, this approach faces the limited resource’s availability in the wireless sensor. The power consumption is considered to be the major concern for large data transmission and extensive processing. Thus, the design of successful audio based solution for target recognition should consider a trade-off between application efficiency and sensor capabilities. The main contribution of this paper is to design a low-power scheme for target detection and recognition based on acoustic signal. This scheme, using features extraction, is intended to locally detect a specific target and to notify a remote se
rver with low energy consumption. This paper details the specification of the proposed scheme and explores its performances for low-power target recognition. The results showed the hypothesis' validity, and demonstrate that the proposed approach can produce classifications as accurate as 96.88% at a very low computational cost.
(More)