Distributed Coordination for Signal Detection in Sensor Networks

假设测试通常被称为信号检测，是信号处理的重要应用之一，在工程和科学中具有重要的意义。虽然信号检测已经研究了很多年，但直到最近才考虑到传感器网络应用的节能信号检测。能源效率似乎是传感器网络最关键的障碍。传感器网络有望解决许多重要问题，包括改善人体、建筑物、桥梁、能源生产设施、森林和其他关键基础设施的监测、控制和修复，同时也为国土安全、执法、灾害预测/避免和国防相关问题提供重要贡献。研究人员打算证明，在没有传感器间通信的情况下，以分布式方式协调多个分散传感器的操作是可能的，这样可以节省大量的能量，而不会影响检测性能。关键是要综合考虑信号检测和通信。这可以通过让具有最多信息观测的传感器首先传输来实现，而传感器或融合中心则收听以跟踪可能假设的总可能性。这使得当一个假设的证据变得压倒性时，传感器传输可以停止。该方法概括了以前的顺序统计和顺序测试方法，同时以一种特定的方式优化性能，该方式为信号处理、通信和网络的高度分布式操作模式建模。该项目的目标是充分理解采用分布式处理的传感器网络的节能信号检测理论。将开发和测试新的方法，重点放在性能和复杂性估计上。

Distributed Coordination for Signal Detection in Sensor NetworksHypothesis testing, often called signal detection, is essential for Engineering and Science while being one of the most important applications of signal processing. Though signal detection has been studied for many years, only recently has energy efficient signal detection been consideredfor sensor networking applications. Energy efficiency appears to be the most critical barrier for sensor networking. Sensor networks hold great promise for solving many important problems including improved monitoring, control and repair of the human body, buildings, bridges, energy production facilities, forests and other critical infrastructure, while also providing important contributions to homeland security, law enforcement, disaster prediction/avoidance and defense related problems.The investigators intend to demonstrate it is possible to coordinate the operation of multiple dispersed sensors in a distributed manner, without inter-sensor communication, such that significant energy is saved without any loss in detection performance. The key is to jointly consider the signal detection and the communications. This can be achieved by having the sensors with themost informative observations transmit first, while either the sensors or the fusion center listen to keep track of the total likelihood of the possible hypotheses. This allows the sensor transmissions to be halted when the evidence for one hypothesis becomes overwhelming. The approach generalizes previous order statistic and sequential testing approaches while optimizing performance in a specific way that models a highly distributed mode of operation for the signal processing, communication, and networking. The goal of this project is to fully understand the theory of energy efficient signal detection for sensor networks employing distributed processing. New approacheswill be developed and tested with an emphasis on performance and complexity estimation.