Visa Koivunen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Visa Koivunen has a total of 60 co-authors. Here are the top co-authors by number of shared publications:
- Robin Rajamäki
Department of Signal Processing and Acoustics Aalto University Espoo Finland
6 shared publications - Topi Halme
Department of Information and Communications Engineering, Aalto University, 02150 Espoo, Finland
6 shared publications - Eyal Nitzan
School of Electrical and Computer Engineering, Ben-Gurion University of the Negev, Beer-Sheva, Israel
4 shared publications - Mikko Valkama
Tampere Wireless Research Centre, Tampere University, Tampere, Finland
4 shared publications - Henk Wymeersch
Chalmers University of Technology
3 shared publications - Musa Furkan Keskin
Chalmers University of Technology
3 shared publications
Latest Publications
- Open Access
Multi-Stream Quickest Change Detection: Foundations and Recent Advances
This paper provides an overview of recent developments in quickest change detection (QCD) for high-dimensional multi-sensor systems, with an emphasis on settings involving structural constraints and limited sensing resources. Classical QCD methodologies, while well understood in low-dimensional and fully observed regimes, face significant challenges when extended to modern applications characterized by large-scale data, constrained sampling or communication, and heterogeneous signal structures. We review key approaches for handling high dimensionality, including methods that exploit sparsity, and other forms of signal heterogeneity. Additionally, we discuss sampling constraints, where observations must be selected or acquired sequentially under resource limitations. Multi-stream applications can require making multiple detections, for example when detecting changes separately in different streams. The underlying assumptions on probability models, the types of changes taking place, commonly used decision-making criteria, performance indices, and error types are described. We also briefly discuss the application of machine learning in cases where the underlying probability models are not known, or there is a need to select which sensors should monitor the phenomena because of the large scale of the system.
Open AccessMulti-Stream Quickest Change Detection: Foundations and Recent Advances
This paper provides an overview of recent developments in quickest change detection (QCD) for high-dimensional multi-sensor systems, with an emphasis on settings involving structural constraints and limited sensing resources. Classical QCD methodologies, while well understood in low-dimensional and fully observed regimes, face significant challenges when extended to modern applications characterized by large-scale data, constrained sampling or communication, and heterogeneous signal structures. We review key approaches for handling high dimensionality, including methods that exploit sparsity, and other forms of signal heterogeneity. Additionally, we discuss sampling constraints, where observations must be selected or acquired sequentially under resource limitations. Multi-stream applications can require making multiple detections, for example when detecting changes separately in different streams. The underlying assumptions on probability models, the types of changes taking place, commonly used decision-making criteria, performance indices, and error types are described. We also briefly discuss the application of machine learning in cases where the underlying probability models are not known, or there is a need to select which sensors should monitor the phenomena because of the large scale of the system.
- Open Access
- 3 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Radar Systems
- Vol. 4, 829-840
Reduced Cyclostationarity MIMO Radar Waveform Synthesis Using Generative Deep Learning
In modern radar applications such as MIMO and cognitive radar, large apertures transmitting a large number of waveforms simultaneously are employed. Waveforms transmitted from different antennas are expected to have desirable properties for radar tasks individually, but also relative to each other to avoid mutual interference. However, existing waveform families are mostly well-known and limited in number of waveforms. In this paper, novel MIMO radar waveform synthesis approaches stemming from generative deep learning are developed. The developed approaches have the potential to discover and generate large numbers of new waveforms on demand with desirable properties. In MIMO radar applications, such properties include close to ideal ambiguity function (AF), constant modulus, and low cross-correlation or orthogonality among the waveforms. In some cases, a particular correlation structure is required to synthesize an arbitrary beampattern. In this paper, a regularized objective for Generative Adversarial Network (GAN)-based learning is introduced to generate Low Probability of Intercept/Detection (LPI/LPD) waveforms with reduced cyclostationarity, while preserving the desirable properties for radar tasks. The capability to generate new waveforms on demand makes it more difficult for the adversary to detect or classify the waveforms employed while the performance of the radar system is not impacted. Moreover, a method where the phase of the training data samples may be dithered to reduce cyclostationarity over the code period is presented. The simulation results demonstrate that the synthesized MIMO radar waveforms possess the desired properties with high fidelity while having improved LPI/LPD properties because of reduced cyclostationarity.
- 3 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Radar Systems
- Vol. 4, 829-840
Open AccessReduced Cyclostationarity MIMO Radar Waveform Synthesis Using Generative Deep Learning
In modern radar applications such as MIMO and cognitive radar, large apertures transmitting a large number of waveforms simultaneously are employed. Waveforms transmitted from different antennas are expected to have desirable properties for radar tasks individually, but also relative to each other to avoid mutual interference. However, existing waveform families are mostly well-known and limited in number of waveforms. In this paper, novel MIMO radar waveform synthesis approaches stemming from generative deep learning are developed. The developed approaches have the potential to discover and generate large numbers of new waveforms on demand with desirable properties. In MIMO radar applications, such properties include close to ideal ambiguity function (AF), constant modulus, and low cross-correlation or orthogonality among the waveforms. In some cases, a particular correlation structure is required to synthesize an arbitrary beampattern. In this paper, a regularized objective for Generative Adversarial Network (GAN)-based learning is introduced to generate Low Probability of Intercept/Detection (LPI/LPD) waveforms with reduced cyclostationarity, while preserving the desirable properties for radar tasks. The capability to generate new waveforms on demand makes it more difficult for the adversary to detect or classify the waveforms employed while the performance of the radar system is not impacted. Moreover, a method where the phase of the training data samples may be dithered to reduce cyclostationarity over the code period is presented. The simulation results demonstrate that the synthesized MIMO radar waveforms possess the desired properties with high fidelity while having improved LPI/LPD properties because of reduced cyclostationarity.
- Open Access
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3
Integrated Radar Sensing and Communications: convergence, co-design and learning
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3
Open AccessIntegrated Radar Sensing and Communications: convergence, co-design and learning
- Open Access
- 1 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 73 (1053587X) , 3963-3976
Downlink Channel Sounding For Multiuser Massive MIMO With Limited Feedback
Massive multiple-input multiple-output (MIMO) technology is pivotal for next-generation wireless networks, offering higher data rates and seamless connectivity. Despite its promise, channel estimation in feedback-based massive MIMO presents unique challenges that must be addressed to realize massive MIMO for frequency-division duplexing (FDD) and other communications systems without channel reciprocity. This paper proposes a novel base station (BS) centric closed-loop channel sounding method for multiuser massive MIMO systems, aiming to reduce the pilot overhead and improve the data rates. It also reduces the computational load on the user equipments (UEs) and facilitates accurate channel estimation at the BS. The proposed approach utilizes beamspace processing in the pilot optimization and the feedback scheme. Furthermore, we show that the proposed method can be employed with hybrid analog-digital architecture in BSs. An efficient algorithm for designing the pilots to maximize the approximate mutual information (MI) between the channel coefficients and received feedback signals is developed, along with an efficient hybrid zero-forcing (HZF) precoding algorithm. The simulation results show that this approach significantly improves channel estimation accuracy and sum rate performance in multiuser massive MIMO systems.
- 1 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 73 (1053587X) , 3963-3976
Open AccessDownlink Channel Sounding For Multiuser Massive MIMO With Limited Feedback
Massive multiple-input multiple-output (MIMO) technology is pivotal for next-generation wireless networks, offering higher data rates and seamless connectivity. Despite its promise, channel estimation in feedback-based massive MIMO presents unique challenges that must be addressed to realize massive MIMO for frequency-division duplexing (FDD) and other communications systems without channel reciprocity. This paper proposes a novel base station (BS) centric closed-loop channel sounding method for multiuser massive MIMO systems, aiming to reduce the pilot overhead and improve the data rates. It also reduces the computational load on the user equipments (UEs) and facilitates accurate channel estimation at the BS. The proposed approach utilizes beamspace processing in the pilot optimization and the feedback scheme. Furthermore, we show that the proposed method can be employed with hybrid analog-digital architecture in BSs. An efficient algorithm for designing the pilots to maximize the approximate mutual information (MI) between the channel coefficients and received feedback signals is developed, along with an efficient hybrid zero-forcing (HZF) precoding algorithm. The simulation results show that this approach significantly improves channel estimation accuracy and sum rate performance in multiuser massive MIMO systems.
- Open Access
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Signal Processing
- Vol. 19 (6) , 1159-1171
One-Bit Direction-of-Arrival Estimation via AdaBoost
The use of one-bit analog-to-digital converters (ADCs) for the direction-of-arrival (DOA) estimation problem offers several advantages over high-precision counterparts, such as reductions in power consumption and production costs. However, it also introduces new challenges, including information loss due to discarding of amplitude information and preservation of signal sign only. As a result, customized DOA estimation methods need to be developed. In this work, two novel learning-based one-bit DOA estimation algorithms are proposed. The proposed learning-based algorithms integrate a specific approximate Gaussian discriminant analysis (GDA) as weak classifiers within an adaptive boosting (AdaBoost) framework. The first proposed method, referred to as on-grid AdaBoost (OG-AdaBoost), first discretizes the DOA angular interval and builds an over-complete dictionary for the array steering matrix and then estimates the corresponding source signal matrix. The second proposed method, named as MUSIC-AdaBoost, estimates the array steering matrix and source signal matrix in an alternating manner with a constraint on the number of iterations using MUSIC and AdaBoost. Numerical simulations demonstrate the advantages and disadvantages of each proposed method compared to state-of-the-art in various challenging scenarios. The results highlight that OG-AdaBoost performs effectively in scenarios involving widely spaced DOAs. Furthermore, the numerical results demonstrate the MUSIC-AdaBoost method's superior performance relative to state-of-the-art algorithms in the face of closely spaced and/or highly correlated sources when both the number of snapshots and the number of antennas are greater than or equal to 40.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Signal Processing
- Vol. 19 (6) , 1159-1171
Open AccessOne-Bit Direction-of-Arrival Estimation via AdaBoost
The use of one-bit analog-to-digital converters (ADCs) for the direction-of-arrival (DOA) estimation problem offers several advantages over high-precision counterparts, such as reductions in power consumption and production costs. However, it also introduces new challenges, including information loss due to discarding of amplitude information and preservation of signal sign only. As a result, customized DOA estimation methods need to be developed. In this work, two novel learning-based one-bit DOA estimation algorithms are proposed. The proposed learning-based algorithms integrate a specific approximate Gaussian discriminant analysis (GDA) as weak classifiers within an adaptive boosting (AdaBoost) framework. The first proposed method, referred to as on-grid AdaBoost (OG-AdaBoost), first discretizes the DOA angular interval and builds an over-complete dictionary for the array steering matrix and then estimates the corresponding source signal matrix. The second proposed method, named as MUSIC-AdaBoost, estimates the array steering matrix and source signal matrix in an alternating manner with a constraint on the number of iterations using MUSIC and AdaBoost. Numerical simulations demonstrate the advantages and disadvantages of each proposed method compared to state-of-the-art in various challenging scenarios. The results highlight that OG-AdaBoost performs effectively in scenarios involving widely spaced DOAs. Furthermore, the numerical results demonstrate the MUSIC-AdaBoost method's superior performance relative to state-of-the-art algorithms in the face of closely spaced and/or highly correlated sources when both the number of snapshots and the number of antennas are greater than or equal to 40.
- Open Access
- 14 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 71 (10) , 7802-7814
Quickest Change Detection for Multiple Data Streams Using the James–Stein Estimator
The problem of quickest change detection is studied in the context of detecting an arbitrary unknown mean-shift in multiple independent Gaussian data streams. The James-Stein estimator is used in constructing detection schemes that exhibit strong detection performance both asymptotically and non-asymptotically. Our results indicate that utilizing the James-Stein estimator in the recently developed window-limited CuSum test constitutes a uniform improvement over its typical maximum likelihood variant. That is, the proposed James-Stein version achieves a smaller detection delay simultaneously for all possible post-change parameter values and every false alarm rate constraint, as long as the number of parallel data streams is greater than three. Additionally, an alternative detection procedure that utilizes the James-Stein estimator is shown to have asymptotic detection delay properties that compare favorably to existing tests. The second-order asymptotic detection delay term is reduced in a predefined low-dimensional subspace of the parameter space, while second-order asymptotic minimaxity is preserved. The results are verified in simulations, where the proposed schemes are shown to achieve smaller detection delays compared to existing alternatives, especially when the number of data streams is large.
- 14 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 71 (10) , 7802-7814
Open AccessQuickest Change Detection for Multiple Data Streams Using the James–Stein Estimator
The problem of quickest change detection is studied in the context of detecting an arbitrary unknown mean-shift in multiple independent Gaussian data streams. The James-Stein estimator is used in constructing detection schemes that exhibit strong detection performance both asymptotically and non-asymptotically. Our results indicate that utilizing the James-Stein estimator in the recently developed window-limited CuSum test constitutes a uniform improvement over its typical maximum likelihood variant. That is, the proposed James-Stein version achieves a smaller detection delay simultaneously for all possible post-change parameter values and every false alarm rate constraint, as long as the number of parallel data streams is greater than three. Additionally, an alternative detection procedure that utilizes the James-Stein estimator is shown to have asymptotic detection delay properties that compare favorably to existing tests. The second-order asymptotic detection delay term is reduced in a predefined low-dimensional subspace of the parameter space, while second-order asymptotic minimaxity is preserved. The results are verified in simulations, where the proposed schemes are shown to achieve smaller detection delays compared to existing alternatives, especially when the number of data streams is large.
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 73 (1053587X) , 2464-2479
The Adaptive $\tau$-Lasso: Robustness and Oracle Properties
This paper introduces a new regularized version of the robust τ-regression estimator for analyzing high-dimensional datasets subject to gross contamination in the response variables and covariates (explanatory variables). The resulting estimator, termed adaptive τ-Lasso, is robust to outliers and high-leverage points. It also incorporates an adaptive l1-norm penalty term, which enables the selection of relevant variables and reduces the bias associated with large true regression coefficients. More specifically, this adaptive l1-norm penalty term assigns a weight to each regression coefficient. For a fixed number of predictors p, we show that the adaptive τ-Lasso has the oracle property, ensuring both variable-selection consistency and asymptotic normality under fairly mild conditions. Asymptotic normality applies only to the entries of the regression vector corresponding to the true support, assuming knowledge of the true regression vector support. We characterize its robustness by establishing the finite-sample breakdown point and the influence function. We carry out extensive simulations and observe that the class of τ-Lasso estimators exhibits robustness and reliable performance in both contaminated and uncontaminated data settings. We also validate our theoretical findings on robustness properties through simulations. In the face of outliers and high-leverage points, the adaptive τ-Lasso and τ-Lasso estimators achieve the best performance or match the best performances of competing regularized estimators, with minimal or no loss in terms of prediction and variable selection accuracy for almost all scenarios considered in this study. Therefore, the adaptive τ-Lasso and τ-Lasso estimators provide attractive tools for a variety of sparse linear regression problems, particularly in high-dimensional settings and when the data is contaminated by outliers and high-leverage points. However, it is worth noting that no particular estimator uniformly dominates others in all considered scenarios
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 73 (1053587X) , 2464-2479
The Adaptive $\tau$-Lasso: Robustness and Oracle Properties
This paper introduces a new regularized version of the robust τ-regression estimator for analyzing high-dimensional datasets subject to gross contamination in the response variables and covariates (explanatory variables). The resulting estimator, termed adaptive τ-Lasso, is robust to outliers and high-leverage points. It also incorporates an adaptive l1-norm penalty term, which enables the selection of relevant variables and reduces the bias associated with large true regression coefficients. More specifically, this adaptive l1-norm penalty term assigns a weight to each regression coefficient. For a fixed number of predictors p, we show that the adaptive τ-Lasso has the oracle property, ensuring both variable-selection consistency and asymptotic normality under fairly mild conditions. Asymptotic normality applies only to the entries of the regression vector corresponding to the true support, assuming knowledge of the true regression vector support. We characterize its robustness by establishing the finite-sample breakdown point and the influence function. We carry out extensive simulations and observe that the class of τ-Lasso estimators exhibits robustness and reliable performance in both contaminated and uncontaminated data settings. We also validate our theoretical findings on robustness properties through simulations. In the face of outliers and high-leverage points, the adaptive τ-Lasso and τ-Lasso estimators achieve the best performance or match the best performances of competing regularized estimators, with minimal or no loss in terms of prediction and variable selection accuracy for almost all scenarios considered in this study. Therefore, the adaptive τ-Lasso and τ-Lasso estimators provide attractive tools for a variety of sparse linear regression problems, particularly in high-dimensional settings and when the data is contaminated by outliers and high-leverage points. However, it is worth noting that no particular estimator uniformly dominates others in all considered scenarios
- 28 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 72 (1053587X) , 5604-5615
Causal Influence in Federated Edge Inference
In this paper, we consider a setting where heterogeneous agents with connectivity are performing inference using unlabeled streaming data. Observed data are only partially informative about the target variable of interest. In order to overcome the uncertainty, agents cooperate with each other by exchanging their local inferences with and through a fusion center. To evaluate how each agent influences the overall decision, we adopt a causal framework in order to distinguish the actual influence of agents from mere correlations within the decision-making process. Various scenarios reflecting different agent participation patterns and fusion center policies are investigated. We derive expressions to quantify the causal impact of each agent on the joint decision, which could be beneficial for anticipating and addressing atypical scenarios, such as adversarial attacks or system malfunctions. We validate our theoretical results with numerical simulations and a real-world application of multicamera crowd counting.
- 28 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 72 (1053587X) , 5604-5615
Causal Influence in Federated Edge Inference
In this paper, we consider a setting where heterogeneous agents with connectivity are performing inference using unlabeled streaming data. Observed data are only partially informative about the target variable of interest. In order to overcome the uncertainty, agents cooperate with each other by exchanging their local inferences with and through a fusion center. To evaluate how each agent influences the overall decision, we adopt a causal framework in order to distinguish the actual influence of agents from mere correlations within the decision-making process. Various scenarios reflecting different agent participation patterns and fusion center policies are investigated. We derive expressions to quantify the causal impact of each agent on the joint decision, which could be beneficial for anticipating and addressing atypical scenarios, such as adversarial attacks or system malfunctions. We validate our theoretical results with numerical simulations and a real-world application of multicamera crowd counting.
- Top Cited Papers
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Magazine
- Vol. 41 (5) , 17-30
Multicarrier ISAC: Advances in waveform design, signal processing, and learning under nonidealities
This paper addresses the topic of integrated sensing and communications (ISAC) in 5G and emerging 6G wireless networks. ISAC systems operate within shared, congested or even contested spectrum, aiming to deliver high performance in both wireless communications and radio frequency (RF) sensing. The expected benefits include more efficient utilization of spectrum, power, hardware (HW) and antenna resources. Focusing on multicarrier (MC) systems, which represent the most widely used communication waveforms, it explores the co-design and optimization of waveforms alongside multiantenna transceiver signal processing for communications and both monostatic and bistatic sensing applications of ISAC. Moreover, techniques of high practical relevance for overcoming and even harnessing challenges posed by non-idealities in actual transceiver implementations are considered. To operate in highly dynamic radio environments and target scenarios, both model-based structured optimization and learning-based methodologies for ISAC systems are covered, assessing their adaptability and learning capabilities under real-world conditions. The paper presents trade-offs in communication-centric and radar-sensing-centric approaches, aiming for an optimized balance in densely used spectrum.
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Magazine
- Vol. 41 (5) , 17-30
Top Cited PapersMulticarrier ISAC: Advances in waveform design, signal processing, and learning under nonidealities
This paper addresses the topic of integrated sensing and communications (ISAC) in 5G and emerging 6G wireless networks. ISAC systems operate within shared, congested or even contested spectrum, aiming to deliver high performance in both wireless communications and radio frequency (RF) sensing. The expected benefits include more efficient utilization of spectrum, power, hardware (HW) and antenna resources. Focusing on multicarrier (MC) systems, which represent the most widely used communication waveforms, it explores the co-design and optimization of waveforms alongside multiantenna transceiver signal processing for communications and both monostatic and bistatic sensing applications of ISAC. Moreover, techniques of high practical relevance for overcoming and even harnessing challenges posed by non-idealities in actual transceiver implementations are considered. To operate in highly dynamic radio environments and target scenarios, both model-based structured optimization and learning-based methodologies for ISAC systems are covered, assessing their adaptability and learning capabilities under real-world conditions. The paper presents trade-offs in communication-centric and radar-sensing-centric approaches, aiming for an optimized balance in densely used spectrum.
- Open Access
- 13 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 70 (11) , 8180-8203
Communication-Constrained Secret Key Generation: Second-Order Bounds
We study communication-constrained secret key generation, where two legitimate parties would like to generate a secret key using communication subject to a rate constraint. The problem is studied in the finite-blocklength regime. In this regime, the use of auxiliary random variables subject to Markov chain conditions in the corresponding asymptotic bounds has proven to make most existing proof techniques insufficient. However, two recently proposed proof techniques – one for the achievability side based on Poisson matching, and another for the converse side based on reverse hypercontractivity – allow us to overcome these issues to some extent. Based on these techniques, novel one-shot and second-order achievability and converse bounds are derived for the problem. While the second-order bounds do not coincide, leaving a precise second-order characterization of the problem an open issue, they improve upon the previously known tightest bounds. The second-order bounds are demonstrated for two simple sources: the binary symmetric source and the Gaussian symmetric source. For the binary source, we find that the gap between the two bounds is mainly due to an unwanted constant in the converse bound, and the non-convexity of the achievability bound.
- 13 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 70 (11) , 8180-8203
Open AccessCommunication-Constrained Secret Key Generation: Second-Order Bounds
We study communication-constrained secret key generation, where two legitimate parties would like to generate a secret key using communication subject to a rate constraint. The problem is studied in the finite-blocklength regime. In this regime, the use of auxiliary random variables subject to Markov chain conditions in the corresponding asymptotic bounds has proven to make most existing proof techniques insufficient. However, two recently proposed proof techniques – one for the achievability side based on Poisson matching, and another for the converse side based on reverse hypercontractivity – allow us to overcome these issues to some extent. Based on these techniques, novel one-shot and second-order achievability and converse bounds are derived for the problem. While the second-order bounds do not coincide, leaving a precise second-order characterization of the problem an open issue, they improve upon the previously known tightest bounds. The second-order bounds are demonstrated for two simple sources: the binary symmetric source and the Gaussian symmetric source. For the binary source, we find that the gap between the two bounds is mainly due to an unwanted constant in the converse bound, and the non-convexity of the achievability bound.