Audio-on-Demand Over Wireless Sensor Networks

Audio represents one of the most appealing yet least exploited modalities inwireless sensor networks, due to the contradiction between the potentially extremelylarge audio data volumes and limited wireless capacity. In this thesis we propose a newparadigm of audio information collection based on the concept of audio-on-demand. Weconsider a infrastructure-free environment targeting post-earthquake rescue and proposean audio-on-demand system over wireless sensor networks.The system needs to search audio data when providing audio-on-demand service.The critical challenge of this thesis is how to guarantee a high search success rate withoutinfrastructure support and at the same time as well as consume as less energy of thesystem as possible. To address the challenge, we propose the probabilistic exhaustivesearch strategy and communication cost minimizing principle, then design our systemmodel. We prove the optimality of the energy consumption of the algorithm with theguarantee of a high search success rate. Based on our system model, we design a novelreplication algorithm that deploys an optimal number of replicas across the sensornetwork, and a data query search scheme corresponding to the replication algorithm. Tofurther reduce the communication cost for the replication, we utilize bloom filter tocompress the metadata of the raw audio data chunks residing on a sensor node.We implement the SAoD system with30IRIS motes equipped with TinyOSoperating system as a real test-bed and also propose several metrics to evaluate theperformance of the system. The experimental results show that our design can providesatisfactory quality of audio-on-demand service with short startup latency and slightplayback jitter. We also conduct extensive simulations to evaluate the efficiency of ourdesign. The extensive simulation results show that our design achieves a search successrate of98%while reducing the search energy consumption by an order of magnitudecompared with existing schemes. Our design also represents better performance ofprecision and scalability.