Volkan Cevher



Theory and Methods For Model-Based Compressive Sensing

ONR Award N00014-08-1-1112, 2008-2012


This project investigates three closely inter-related research themes to develop:

  • New model-based signal recovery algorithms that significantly reduce the number of measurements and computations required for stable recovery as compared to the state-of-the-art.

  • New model-based compressive inference algorithms that extract the key signal information from CS measurements without first reconstructing the signal.

  • Theoretical foundations of model-based CS that both provide performance bounds for model-based CS and suggest approaches to adapt current systems to a model-based setting.

Information Scalable Analog-to-Information Receivers

DARPA Award N66001-08-1-2065, 2008-2011


This project develops an analog-to-information receiver network to enable demodulation and localization of wireless sources that are hidden in bandwidths greater than 250GHz.


Decentralized ISR and Comms for Wireless Sensor Networks

ARL Award DAAD19-01-02-0008, MS-06-37, 2005-2006

The project focused on decentralized ISR (intelligence, surveillance, and reconnaissance) and communications for wireless networks. It was a joint research effort in sensing, communications, and advanced decision support for sensor networks connected by ad-hoc wireless communication links. The project investigated joint sensor signal processing and data fusion design, communication design, and human decision-making in sensor network applications.

Sensor Networks: Signal Processing and Computer Vision

Multi modal, multi sensor fusion for detection, classification and tracking of humans/vehicles

ARL Award DAAD19-01-02-0008, MS-07-08A, 2006-2007

Decentralized data fusion for multi-sensor multi-mode multi-target detection, classification, and tracking

ARL Award DAAD19-01-02-0008, MS-06-01, 2005-2006


Acoustic sensor network design for position estimation project developed a theory to predict the localization performance of randomly distributed sensor networks consisting of various sensor modalities when only a constant active subset of sensors that minimize localization error is used for estimation. The characteristics of the modalities include measurement type (bearing or range) and error, sensor reliability, FOV, sensing range, and mobility. The project showed that the localization performance of a sensor network is a function of a weighted sum of the total number of each sensor modality; this weighted sum is independent of how the sensor management strategy chooses the active sensors. The project also combined the utility objective with other objectives, such as lifetime, coverage and reliability to determine the best mix of sensors for an optimal sensor network design: the Pareto efficient frontier.

Joint acoustic-video fingerprinting of vehicles project addressed vehicle classification and mensuration problems using acoustic and video sensors. In the resulting publications, we showed how to estimate a vehicle’s speed, width, and length by jointly estimating its acoustic wave-pattern using a single passive acoustic sensor that records the vehicle’s drive-by noise. We also showed joint fusion results with video using vehicle classification and mensuration based on silhouette extraction and wheel detection. Vehicle speed estimation and classification results are provided using field data.

Joint acoustic-video object tracking project resulted in a Bayesian framework for target tracking using multi modal information. The project developed a joint acoustic video algorithm using particle filters for detecting and tracking multiple vehicles and also successfully extended the Bayesian framework for joint acoustic radar and radar seismic algorithms.