This paper presents the key ideas behind collaborative signal processing (CSP) algorithms for distributed sensor networks developed at the University of Wisconsin (UW). The authors describe how these algorithms interface with the UW-API networking/routing algorithms. The framework is motivated by the problem of detecting and tracking a single maneuvering target, illustrating the integration between UW-API and UW-CSP algorithms. The paper also discusses the challenges of tracking multiple targets, which requires classification techniques. The authors consider the signal processing aspects of this problem under the constraints imposed by limited node capabilities and networking/routing constraints. The paper discusses the performance versus complexity trade-off in sensor networks, where collaboration between nodes comes at the cost of exchanging information. The authors focus on target classification, which is arguably the most challenging signal processing task in sensor networks. They provide numerical results based on real data that offer insights into the problem and help identify key issues and challenges. The paper also identifies promising directions for future research. The paper discusses the key components of the CSP framework for tracking multiple targets in a distributed sensor network, including event detection, estimation and prediction of target location, and target classification. The authors argue that spectral target signatures may yield accurate classification, but variations in spectral signatures due to various effects, including Doppler shifts, present a significant challenge. The paper presents preliminary results on classifying between wheeled and tracked vehicles, with the SVM classifier seeming to be the most promising based on initial experiments. The paper also discusses the challenges of making CSP work in real sensor networks, including uncertainty in temporal and spatial measurements, Doppler effects due to motion, and variability in experimental conditions. The authors emphasize the importance of making classifiers robust to such variability for their successful application. The paper concludes by highlighting the emerging potential of distributed sensor networks as a powerful technology for remotely instrumenting and monitoring the physical world, while acknowledging that many challenges need to be overcome before it becomes practically feasible. The authors also mention that they are working on incorporating insights and results obtained by other researchers and that results from this ongoing project, including code for implementing various algorithms, will be posted on their website.This paper presents the key ideas behind collaborative signal processing (CSP) algorithms for distributed sensor networks developed at the University of Wisconsin (UW). The authors describe how these algorithms interface with the UW-API networking/routing algorithms. The framework is motivated by the problem of detecting and tracking a single maneuvering target, illustrating the integration between UW-API and UW-CSP algorithms. The paper also discusses the challenges of tracking multiple targets, which requires classification techniques. The authors consider the signal processing aspects of this problem under the constraints imposed by limited node capabilities and networking/routing constraints. The paper discusses the performance versus complexity trade-off in sensor networks, where collaboration between nodes comes at the cost of exchanging information. The authors focus on target classification, which is arguably the most challenging signal processing task in sensor networks. They provide numerical results based on real data that offer insights into the problem and help identify key issues and challenges. The paper also identifies promising directions for future research. The paper discusses the key components of the CSP framework for tracking multiple targets in a distributed sensor network, including event detection, estimation and prediction of target location, and target classification. The authors argue that spectral target signatures may yield accurate classification, but variations in spectral signatures due to various effects, including Doppler shifts, present a significant challenge. The paper presents preliminary results on classifying between wheeled and tracked vehicles, with the SVM classifier seeming to be the most promising based on initial experiments. The paper also discusses the challenges of making CSP work in real sensor networks, including uncertainty in temporal and spatial measurements, Doppler effects due to motion, and variability in experimental conditions. The authors emphasize the importance of making classifiers robust to such variability for their successful application. The paper concludes by highlighting the emerging potential of distributed sensor networks as a powerful technology for remotely instrumenting and monitoring the physical world, while acknowledging that many challenges need to be overcome before it becomes practically feasible. The authors also mention that they are working on incorporating insights and results obtained by other researchers and that results from this ongoing project, including code for implementing various algorithms, will be posted on their website.