Skip to main content

Marco Duarte

University of Massachusetts Amherst, Electrical and Computer Engineering, Faculty Member

Followers

104

Following

29

Public Views

Natallia Katenka

University of Rhode Island

Aleksandar Dogandžić

Iowa State University

sundeep prabhakar C

Delft University of Technology

Israel Josafat Perez

Instituto Politécnico Nacional

prajakta kshirsagar

Annamalai Annamalai

buthina Qatatsheh

McNeese State University

Interests

Uploads

Papers by Marco Duarte

Collaborative signal processing in distributed wireless sensor networks

The Second International Workshop on Information …, Jan 1, 2003

Distributed wireless sensor networks have been proposed as a solution to environment sensing, tar... more Distributed wireless sensor networks have been proposed as a solution to environment sensing, target tracking, data collection and others. These problems involve estimation or detection problems using data collected by individual sensors. Due to the characteristics of the DWSN, it is desired to use algorithms that will allow for low-cost information communication and fusion. Some of the recent approaches to these problems are detailed in this paper. There is a heavy focus on distributed estimation and optimization algorithms that feature lower power consumption and communication requirements than the standard centralized architecture.

A network API-driven survey of communication requirements of distributed data processing algorithms for sensor networks

In , we present the design of a sensor network application programming interface (API) to cover t... more In , we present the design of a sensor network application programming interface (API) to cover the common communication requirements of data processing applications in sensor networks, motivated by a study of all such applications proposed to date in the proceedings of Information Processing in Sensor Networks (IPSN). Due to space limitations, only a subset of the more than 100 papers surveyed are discussed in . In this technical report, we present the full results of the survey.

Recovery of compressible signals in unions of subspaces

… Sciences and Systems, …, Jan 1, 2009

Public reporting burden for the collection of information is estimated to average 1 hour per resp... more Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.

Recovery of frequency-sparse signals from compressive measurements

Communication, Control, and …, Jan 1, 2010

Compressive sensing (CS) is a new approach to simultaneous sensing and compression for sparse and... more Compressive sensing (CS) is a new approach to simultaneous sensing and compression for sparse and compressible signals. While the discrete Fourier transform has been widely used for CS of frequency-sparse signals, it provides optimal sparse representations only for signals with components at integral frequencies. There exist redundant frames that provide compressible representations for frequency-sparse signals, but such frames are highly coherent and severely affect the performance of standard CS recovery. In this paper, we show that by modifying standard CS recovery algorithms to prevent coherent frame elements from being present in the signal estimate, it is possible to bypass the shortcomings introduced by the coherent frame. The resulting algorithm comes with theoretical guarantees and is shown to perform significantly better for frequencysparse signal recovery than its standard counterparts. The algorithm can also be extended to similar settings that use coherent frames.

Fast reconstruction from random incoherent projections

The Compressed Sensing paradigm consists of recovering signals that can be sparsely represented i... more The Compressed Sensing paradigm consists of recovering signals that can be sparsely represented in a given basis from a small set of projections into random vectors; the problem is typically solved using an adapted Basis Pursuit algorithm. We show that the recovery of the signal is equivalent to determining the sparsest representation of the measurement vector using the dictionary obtained by applying the projections to the basis elements, and therefore more efficient algorithms can be used during recovery; we explore the Matching Pursuit and Orthogonal Matching Pursuit algorithms. We also design an algorithm that allows for even faster recovery for piecewise smooth signals: the algorithm exploits the tree structure of the sparse coefficients of the signal in a wavelet representation to select subsets of coefficients to estimate at each iteration. We define a class of signals for which such a Tree Matching Pursuit algorithm performs successful recovery and present variations of the algorithms for different classes of signals. Other applications of TMP include effective denoising, fast approximation in overcomplete wavelet bases, and distributed compression.

Texas Hold'Em algorithms for distributed compressive sensing

… Speech and Signal …, Jan 1, 2010

This paper develops a new class of algorithms for signal recovery in the distributed compressive ... more This paper develops a new class of algorithms for signal recovery in the distributed compressive sensing (DCS) framework. DCS exploits both intra-signal and inter-signal correlations through the concept of joint sparsity to further reduce the number of measurements required for recovery. DCS is wellsuited for sensor network applications due to its universality, computational asymmetry, tolerance to quantization and noise, and robustness to measurement loss. In this paper we propose recovery algorithms for the sparse common and innovation joint sparsity model. Our approach leads to a class of efficient algorithms, the Texas Hold 'Em algorithms, which are scalable both in terms of communication bandwidth and computational complexity.

A network application programming interface for data processing in sensor networks

Submitted to …, Jan 1, 2007

Since the inception of sensor networks, a wide variety of algorithms for in-network data processi... more Since the inception of sensor networks, a wide variety of algorithms for in-network data processing have emerged. To enable practical implementation of a broad class of these proposed algorithms, sufficient application programming support for network communications is critical. While standard (but relatively low-level) network application programming interfaces (APIs) such as those implemented in TinyOS can provide a basis for flexible application-specific customization, experience in the past several years has shown that data processing algorithms in fact share similar higher level communication needs that can be better supported.

Multiuser detection in asynchronous on-off random access channels using lasso

… (Allerton), 2010 48th …, Jan 1, 2010

This paper considers on-off random access channels where users transmit either a one or a zero to... more This paper considers on-off random access channels where users transmit either a one or a zero to a base station. Such channels represent an abstraction of control channels used for scheduling requests in third-generation cellular systems and uplinks in wireless sensor networks deployed for target detection. This paper introduces a novel convex-optimization-based scheme for multiuser detection (MUD) in asynchronous on-off random access channels that does not require knowledge of the delays or the instantaneous received signal-to-noise ratios of the individual users at the base station. For any fixed number of temporal signal space dimensions N and maximum delay τ in the system, the proposed scheme can accommodate M exp(O(N 1/3 )) total users and k N/ log M active users in the system-a significant improvement over the k ≤ M N scaling suggested by the use of classical matched-filtering-based approaches to MUD employing orthogonal signaling. Furthermore, the computational complexity of the proposed scheme differs from that of a similar oracle-based scheme with perfect knowledge of the user delays by at most a factor of log(N +τ ). Finally, the results presented in here are nonasymptotic, in contrast to related previous work for synchronous channels that only guarantees that the probability of MUD error at the base station goes to zero asymptotically in M .

Compressive sensing for signal ensembles

The quality of this reproduction is dependent upon the quality of the copy submitted.

Optimal decision fusion for sensor network applications

First ACM Conference on Embedded …, Jan 1, 2003

Energy conservation is a major concern of large scale sensor network systems when deployed for lo... more Energy conservation is a major concern of large scale sensor network systems when deployed for long-term monitoring purposes. In such a wireless sensor network, information obtained by individual sensors must be aggregated via wireless communication channels, and hence will consume a considerable amount of energy. This problem can be mitigated using "data fusion" where data sampled at individual sensors is processed locally to calculate a local statistic. These locally processed results then will be transmitted to a "fusion center" via wireless channel to deduce a global statistic. There are two types of data fusion modes: decision fusion where a decision from small discrete set of possible decisions is transmitted from each sensor node; and value fusion where a real number is transmitted from each sensor node. Numerous decision fusion methods have been proposed in the past. Yet, few have claimed optimality in terms of maximizing the probability of correct decision of the fused decision. In this paper, we present a simple, yet practical proof to show how to derive an optimal decision rule for a family of decision fusion problems where sufficient amount of training data samples are available. This proof is general in the sense that it is not tied to any specific existing decision fusion algorithms. Yet, it can be used to access the optimality of any existing decision fusion algorithm for particular set of problems.

Fidelity analysis of mechanically aided copying/enlarging of Jan van Eyck's Portrait of Niccolo Albergati

Proceedings of SPIE, Jan 1, 2007

In this paper we present a computationally efficient Optimal Mass Transport algorithm. This metho... more In this paper we present a computationally efficient Optimal Mass Transport algorithm. This method is based on the Monge-Kantorovich theory and is used for computing elastic registration and warping maps in image registration and morphing applications. This is a parameter free method which utilizes all of the grayscale data in an image pair in a symmetric fashion. No landmarks need to be specified for correspondence. In our work, we demonstrate significant improvement in computation time when our algorithm is applied as compared to the originally proposed method by Haker et al . The original algorithm was based on a gradient descent method for removing the curl from an initial mass preserving map regarded as 2D vector field. This involves inverting the Laplacian in each iteration which is now computed using full multigrid technique resulting in an improvement in computational time by a factor of two. Greater improvement is achieved by decimating the curl in a multi-resolutional framework. The algorithm was applied to 2D short axis cardiac MRI images and brain MRI images for testing and comparison.

Revisiting computer image analysis and art

Multimedia, IEEE, Jan 1, 2007

Did early Renaissance painters trace optically projected images? The conclusion of independent scientists, art historians and artists

Digital Imaging for …, Jan 1, 2011

The performance of group lasso for linear regression of grouped variables

Asynchronous code-division random access using convex optimization

Physical …, Jan 1, 2011

Optimal decision fusion with applications to target detection in wireless ad hoc sensor networks

Multimedia and Expo, 2004. ICME'04. …, Jan 1, 2004

Decision fusion is a decentralized decision making process where local decisions are combined to ... more Decision fusion is a decentralized decision making process where local decisions are combined to reach a global decision. In this work, we propose a complementary optimal decision fusion (CODF) method to the target detection task that arises in wireless ad hoc sensor network signal processing. We conduct extensive comparative study using real world sensor signal data, and observe superior performance of CODF when compared with state-of-the-art decision fusion methods. In addition to distributed sensor network applications, the proposed CODF algorithm can be applied to numerous multi-modality, multi-agent, multi-media signal processing problems.

On the conditioning of random block subdictionaries

Department of Computer …, Jan 1, 2010

A theoretical analysis of joint manifolds

Arxiv preprint arXiv: …, Jan 1, 2009

The emergence of low-cost sensor architectures for diverse modalities has made it possible to dep... more The emergence of low-cost sensor architectures for diverse modalities has made it possible to deploy sensor arrays that capture a single event from a large number of vantage points and using multiple modalities. In many scenarios, these sensors acquire very high-dimensional data such as audio signals, images, and video. To cope with such high-dimensional data, we typically rely on low-dimensional models. Manifold models provide a particularly powerful model that captures the structure of high-dimensional data when it is governed by a low-dimensional set of parameters. However, these models do not typically take into account dependencies among multiple sensors. We thus propose a new joint manifold framework for data ensembles that exploits such dependencies. We show that simple algorithms can exploit the joint manifold structure to improve their performance on standard signal processing applications. Additionally, recent results concerning dimensionality reduction for manifolds enable us to formulate a network-scalable data compression scheme that uses random projections of the sensed data. This scheme efficiently fuses the data from all sensors through the addition of such projections, regardless of the data modalities and dimensions.

Kronecker product matrices for compressive sensing

Acoustics Speech and Signal …, Jan 1, 2010

Compressive sensing (CS) is an emerging approach for acquisition of signals having a sparse or co... more Compressive sensing (CS) is an emerging approach for acquisition of signals having a sparse or compressible representation in some basis. While CS literature has mostly focused on problems involving 1-D and 2-D signals, many important applications involve signals that are multidimensional. We propose the use of Kronecker product matrices in CS for two purposes. First, we can use such matrices as sparsifying bases that jointly model the different types of structure present in the signal. Second, the measurement matrices used in distributed measurement settings can be easily expressed as Kronecker products. This new formulation enables the derivation of analytical bounds for sparse approximation and CS recovery of multidimensional signals.

Joint manifolds for data fusion

… IEEE Transactions on, Jan 1, 2010

The emergence of low-cost sensing architectures for diverse modalities has made it possible to de... more The emergence of low-cost sensing architectures for diverse modalities has made it possible to deploy sensor networks that capture a single event from a large number of vantage points and using multiple modalities. In many scenarios, these networks acquire large amounts of very high-dimensional data. For example, even a relatively small network of cameras can generate massive amounts of high-dimensional image and video data. One way to cope with this data deluge is to exploit low-dimensional data models. Manifold models provide a particularly powerful theoretical and algorithmic framework for capturing the structure of data governed by a small number of parameters, as is often the case in a sensor network. However, these models do not typically take into account dependencies among multiple sensors. We thus propose a new joint manifold framework for data ensembles that exploits such dependencies. We show that joint manifold structure can lead to improved performance for a variety of signal processing algorithms for applications including classification and manifold learning. Additionally, recent results concerning random projections of manifolds enable us to formulate a scalable and universal dimensionality reduction scheme that efficiently fuses the data from all sensors.

Collaborative signal processing in distributed wireless sensor networks

The Second International Workshop on Information …, Jan 1, 2003

Distributed wireless sensor networks have been proposed as a solution to environment sensing, tar... more Distributed wireless sensor networks have been proposed as a solution to environment sensing, target tracking, data collection and others. These problems involve estimation or detection problems using data collected by individual sensors. Due to the characteristics of the DWSN, it is desired to use algorithms that will allow for low-cost information communication and fusion. Some of the recent approaches to these problems are detailed in this paper. There is a heavy focus on distributed estimation and optimization algorithms that feature lower power consumption and communication requirements than the standard centralized architecture.

A network API-driven survey of communication requirements of distributed data processing algorithms for sensor networks

In , we present the design of a sensor network application programming interface (API) to cover t... more In , we present the design of a sensor network application programming interface (API) to cover the common communication requirements of data processing applications in sensor networks, motivated by a study of all such applications proposed to date in the proceedings of Information Processing in Sensor Networks (IPSN). Due to space limitations, only a subset of the more than 100 papers surveyed are discussed in . In this technical report, we present the full results of the survey.

Recovery of compressible signals in unions of subspaces

… Sciences and Systems, …, Jan 1, 2009

Public reporting burden for the collection of information is estimated to average 1 hour per resp... more Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.

Recovery of frequency-sparse signals from compressive measurements

Communication, Control, and …, Jan 1, 2010

Compressive sensing (CS) is a new approach to simultaneous sensing and compression for sparse and... more Compressive sensing (CS) is a new approach to simultaneous sensing and compression for sparse and compressible signals. While the discrete Fourier transform has been widely used for CS of frequency-sparse signals, it provides optimal sparse representations only for signals with components at integral frequencies. There exist redundant frames that provide compressible representations for frequency-sparse signals, but such frames are highly coherent and severely affect the performance of standard CS recovery. In this paper, we show that by modifying standard CS recovery algorithms to prevent coherent frame elements from being present in the signal estimate, it is possible to bypass the shortcomings introduced by the coherent frame. The resulting algorithm comes with theoretical guarantees and is shown to perform significantly better for frequencysparse signal recovery than its standard counterparts. The algorithm can also be extended to similar settings that use coherent frames.

Fast reconstruction from random incoherent projections

The Compressed Sensing paradigm consists of recovering signals that can be sparsely represented i... more The Compressed Sensing paradigm consists of recovering signals that can be sparsely represented in a given basis from a small set of projections into random vectors; the problem is typically solved using an adapted Basis Pursuit algorithm. We show that the recovery of the signal is equivalent to determining the sparsest representation of the measurement vector using the dictionary obtained by applying the projections to the basis elements, and therefore more efficient algorithms can be used during recovery; we explore the Matching Pursuit and Orthogonal Matching Pursuit algorithms. We also design an algorithm that allows for even faster recovery for piecewise smooth signals: the algorithm exploits the tree structure of the sparse coefficients of the signal in a wavelet representation to select subsets of coefficients to estimate at each iteration. We define a class of signals for which such a Tree Matching Pursuit algorithm performs successful recovery and present variations of the algorithms for different classes of signals. Other applications of TMP include effective denoising, fast approximation in overcomplete wavelet bases, and distributed compression.

Texas Hold'Em algorithms for distributed compressive sensing

… Speech and Signal …, Jan 1, 2010

This paper develops a new class of algorithms for signal recovery in the distributed compressive ... more This paper develops a new class of algorithms for signal recovery in the distributed compressive sensing (DCS) framework. DCS exploits both intra-signal and inter-signal correlations through the concept of joint sparsity to further reduce the number of measurements required for recovery. DCS is wellsuited for sensor network applications due to its universality, computational asymmetry, tolerance to quantization and noise, and robustness to measurement loss. In this paper we propose recovery algorithms for the sparse common and innovation joint sparsity model. Our approach leads to a class of efficient algorithms, the Texas Hold 'Em algorithms, which are scalable both in terms of communication bandwidth and computational complexity.

A network application programming interface for data processing in sensor networks

Submitted to …, Jan 1, 2007

Since the inception of sensor networks, a wide variety of algorithms for in-network data processi... more Since the inception of sensor networks, a wide variety of algorithms for in-network data processing have emerged. To enable practical implementation of a broad class of these proposed algorithms, sufficient application programming support for network communications is critical. While standard (but relatively low-level) network application programming interfaces (APIs) such as those implemented in TinyOS can provide a basis for flexible application-specific customization, experience in the past several years has shown that data processing algorithms in fact share similar higher level communication needs that can be better supported.

Multiuser detection in asynchronous on-off random access channels using lasso

… (Allerton), 2010 48th …, Jan 1, 2010

This paper considers on-off random access channels where users transmit either a one or a zero to... more This paper considers on-off random access channels where users transmit either a one or a zero to a base station. Such channels represent an abstraction of control channels used for scheduling requests in third-generation cellular systems and uplinks in wireless sensor networks deployed for target detection. This paper introduces a novel convex-optimization-based scheme for multiuser detection (MUD) in asynchronous on-off random access channels that does not require knowledge of the delays or the instantaneous received signal-to-noise ratios of the individual users at the base station. For any fixed number of temporal signal space dimensions N and maximum delay τ in the system, the proposed scheme can accommodate M exp(O(N 1/3 )) total users and k N/ log M active users in the system-a significant improvement over the k ≤ M N scaling suggested by the use of classical matched-filtering-based approaches to MUD employing orthogonal signaling. Furthermore, the computational complexity of the proposed scheme differs from that of a similar oracle-based scheme with perfect knowledge of the user delays by at most a factor of log(N +τ ). Finally, the results presented in here are nonasymptotic, in contrast to related previous work for synchronous channels that only guarantees that the probability of MUD error at the base station goes to zero asymptotically in M .

Compressive sensing for signal ensembles

The quality of this reproduction is dependent upon the quality of the copy submitted.

Optimal decision fusion for sensor network applications

First ACM Conference on Embedded …, Jan 1, 2003

Energy conservation is a major concern of large scale sensor network systems when deployed for lo... more Energy conservation is a major concern of large scale sensor network systems when deployed for long-term monitoring purposes. In such a wireless sensor network, information obtained by individual sensors must be aggregated via wireless communication channels, and hence will consume a considerable amount of energy. This problem can be mitigated using "data fusion" where data sampled at individual sensors is processed locally to calculate a local statistic. These locally processed results then will be transmitted to a "fusion center" via wireless channel to deduce a global statistic. There are two types of data fusion modes: decision fusion where a decision from small discrete set of possible decisions is transmitted from each sensor node; and value fusion where a real number is transmitted from each sensor node. Numerous decision fusion methods have been proposed in the past. Yet, few have claimed optimality in terms of maximizing the probability of correct decision of the fused decision. In this paper, we present a simple, yet practical proof to show how to derive an optimal decision rule for a family of decision fusion problems where sufficient amount of training data samples are available. This proof is general in the sense that it is not tied to any specific existing decision fusion algorithms. Yet, it can be used to access the optimality of any existing decision fusion algorithm for particular set of problems.

Fidelity analysis of mechanically aided copying/enlarging of Jan van Eyck's Portrait of Niccolo Albergati

Proceedings of SPIE, Jan 1, 2007

In this paper we present a computationally efficient Optimal Mass Transport algorithm. This metho... more In this paper we present a computationally efficient Optimal Mass Transport algorithm. This method is based on the Monge-Kantorovich theory and is used for computing elastic registration and warping maps in image registration and morphing applications. This is a parameter free method which utilizes all of the grayscale data in an image pair in a symmetric fashion. No landmarks need to be specified for correspondence. In our work, we demonstrate significant improvement in computation time when our algorithm is applied as compared to the originally proposed method by Haker et al . The original algorithm was based on a gradient descent method for removing the curl from an initial mass preserving map regarded as 2D vector field. This involves inverting the Laplacian in each iteration which is now computed using full multigrid technique resulting in an improvement in computational time by a factor of two. Greater improvement is achieved by decimating the curl in a multi-resolutional framework. The algorithm was applied to 2D short axis cardiac MRI images and brain MRI images for testing and comparison.

Revisiting computer image analysis and art

Multimedia, IEEE, Jan 1, 2007

Did early Renaissance painters trace optically projected images? The conclusion of independent scientists, art historians and artists

Digital Imaging for …, Jan 1, 2011

The performance of group lasso for linear regression of grouped variables

Asynchronous code-division random access using convex optimization

Physical …, Jan 1, 2011

Optimal decision fusion with applications to target detection in wireless ad hoc sensor networks

Multimedia and Expo, 2004. ICME'04. …, Jan 1, 2004

Decision fusion is a decentralized decision making process where local decisions are combined to ... more Decision fusion is a decentralized decision making process where local decisions are combined to reach a global decision. In this work, we propose a complementary optimal decision fusion (CODF) method to the target detection task that arises in wireless ad hoc sensor network signal processing. We conduct extensive comparative study using real world sensor signal data, and observe superior performance of CODF when compared with state-of-the-art decision fusion methods. In addition to distributed sensor network applications, the proposed CODF algorithm can be applied to numerous multi-modality, multi-agent, multi-media signal processing problems.

On the conditioning of random block subdictionaries

Department of Computer …, Jan 1, 2010

A theoretical analysis of joint manifolds

Arxiv preprint arXiv: …, Jan 1, 2009

The emergence of low-cost sensor architectures for diverse modalities has made it possible to dep... more The emergence of low-cost sensor architectures for diverse modalities has made it possible to deploy sensor arrays that capture a single event from a large number of vantage points and using multiple modalities. In many scenarios, these sensors acquire very high-dimensional data such as audio signals, images, and video. To cope with such high-dimensional data, we typically rely on low-dimensional models. Manifold models provide a particularly powerful model that captures the structure of high-dimensional data when it is governed by a low-dimensional set of parameters. However, these models do not typically take into account dependencies among multiple sensors. We thus propose a new joint manifold framework for data ensembles that exploits such dependencies. We show that simple algorithms can exploit the joint manifold structure to improve their performance on standard signal processing applications. Additionally, recent results concerning dimensionality reduction for manifolds enable us to formulate a network-scalable data compression scheme that uses random projections of the sensed data. This scheme efficiently fuses the data from all sensors through the addition of such projections, regardless of the data modalities and dimensions.

Kronecker product matrices for compressive sensing

Acoustics Speech and Signal …, Jan 1, 2010

Compressive sensing (CS) is an emerging approach for acquisition of signals having a sparse or co... more Compressive sensing (CS) is an emerging approach for acquisition of signals having a sparse or compressible representation in some basis. While CS literature has mostly focused on problems involving 1-D and 2-D signals, many important applications involve signals that are multidimensional. We propose the use of Kronecker product matrices in CS for two purposes. First, we can use such matrices as sparsifying bases that jointly model the different types of structure present in the signal. Second, the measurement matrices used in distributed measurement settings can be easily expressed as Kronecker products. This new formulation enables the derivation of analytical bounds for sparse approximation and CS recovery of multidimensional signals.

Joint manifolds for data fusion

… IEEE Transactions on, Jan 1, 2010

The emergence of low-cost sensing architectures for diverse modalities has made it possible to de... more The emergence of low-cost sensing architectures for diverse modalities has made it possible to deploy sensor networks that capture a single event from a large number of vantage points and using multiple modalities. In many scenarios, these networks acquire large amounts of very high-dimensional data. For example, even a relatively small network of cameras can generate massive amounts of high-dimensional image and video data. One way to cope with this data deluge is to exploit low-dimensional data models. Manifold models provide a particularly powerful theoretical and algorithmic framework for capturing the structure of data governed by a small number of parameters, as is often the case in a sensor network. However, these models do not typically take into account dependencies among multiple sensors. We thus propose a new joint manifold framework for data ensembles that exploits such dependencies. We show that joint manifold structure can lead to improved performance for a variety of signal processing algorithms for applications including classification and manifold learning. Additionally, recent results concerning random projections of manifolds enable us to formulate a scalable and universal dimensionality reduction scheme that efficiently fuses the data from all sensors.