liu.seSearch for publications in DiVA
Change search
Link to record
Permanent link

Direct link
Gustafsson, Fredrik, ProfessorORCID iD iconorcid.org/0000-0003-3270-171X
Publications (10 of 601) Show all publications
Malmström, M., Kullberg, A., Skog, I., Axehill, D. & Gustafsson, F. (2024). Extended Target Tracking Utilizing Machine-Learning Software–With Applications to Animal Classification. IEEE Signal Processing Letters, 31, 376-380
Open this publication in new window or tab >>Extended Target Tracking Utilizing Machine-Learning Software–With Applications to Animal Classification
Show others...
2024 (English)In: IEEE Signal Processing Letters, ISSN 1070-9908, E-ISSN 1558-2361, Vol. 31, p. 376-380Article in journal (Refereed) Published
Abstract [en]

This letter considers the problem of detecting and tracking objects in a sequence of images. The problem is formulated in a filtering framework, using the output of object-detection algorithms as measurements. An extension to the filtering formulation is proposed that incorporates class information from the previous frame to robustify the classification. Further, the properties of the object-detection algorithm are exploited to quantify the uncertainty of the bounding box detection in each frame. The complete filtering method is evaluated on camera trap images of the four large Swedish carnivores, bear, lynx, wolf, and wolverine. The experiments show that the class tracking formulation leads to a more robust classification.

Place, publisher, year, edition, pages
IEEE, 2024
Keywords
Signal processing algorithms;Classification algorithms;Cameras;Target tracking;Filtering algorithms;Standards;Loss measurement;Multi-object tracking;object detection;environmental monitoring;deep learning;Kalman filters
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-201110 (URN)10.1109/LSP.2024.3353165 (DOI)2-s2.0-85182945517 (Scopus ID)
Available from: 2024-02-21 Created: 2024-02-21 Last updated: 2024-02-22
Forsling, R., Gustafsson, F., Sjanic, Z. & Hendeby, G. (2023). Decentralized Data Fusion of Dimension-Reduced Estimates Using Local Information Only. In: 2023 IEEE Aerospace Conference: . Paper presented at IEEE Aerospace Conference, Big Sky, MT, USA, March 4-11, 2023. IEEE
Open this publication in new window or tab >>Decentralized Data Fusion of Dimension-Reduced Estimates Using Local Information Only
2023 (English)In: 2023 IEEE Aerospace Conference, IEEE , 2023Conference paper, Published paper (Refereed)
Abstract [en]

This paper considers fusion of dimension-reduced estimates in a decentralized sensor network. The benefits of a decentralized sensor network include modularity, robustness and flexibility. Moreover, since preprocessed data is exchanged between the agents it allows for reduced communication. Nevertheless, in certain applications the communication load is required to be reduced even further. One way to decrease the communication load is to exchange dimension-reduced estimates instead of full estimates. Previous work on this topic assumes global availability of covariance matrices, an assumption which is not realistic in decentralized applications. Hence, in this paper we consider the problem of deriving dimension-reduced estimates using only local information. The proposed solution is based on an estimate of the information common to the network. This common information estimate is computed locally at each agent by fusion of all information that is either received or transmitted by that agent. It is shown how the common information estimate is utilized for fusion of dimension-reduced estimates using two well-known fusion methods: the Kalman fuser which is optimal under the assumption of uncorrelated estimates, and covariance intersection. One main theoretical result is that the common information estimate allows for a decorrelation procedure such that uncorrelated estimates can be maintained. This property is crucial to be able to use the Kalman fuser without double counting of information. A numerical comparison suggests that the performance degradation of using the common information estimate, compared to having local access to the actual covariance matrices computed by other agents, is relatively small.

Place, publisher, year, edition, pages
IEEE, 2023
Series
IEEE Aerospace Conference Proceedings, ISSN 1095-323X
Keywords
Target tracking; Decentralized data fusion; Dimension-reduced estimates; Multisensor fusion; Distributed Estimation
National Category
Control Engineering Signal Processing
Identifiers
urn:nbn:se:liu:diva-192747 (URN)10.1109/AERO55745.2023.10115967 (DOI)001008282005026 ()9781665490320 (ISBN)9781665490337 (ISBN)
Conference
IEEE Aerospace Conference, Big Sky, MT, USA, March 4-11, 2023
Funder
Vinnova, Industry Competence Center LINK-SIC
Note

Funding: Industry Excellence Center LINK-SIC - Swedish Governmental Agency for Innovation Systems (VINNOVA); Saab AB

Available from: 2023-03-29 Created: 2023-03-29 Last updated: 2023-09-21Bibliographically approved
Zetterqvist, G., Wahledow, E., Sjövik, P., Gustafsson, F. & Hendeby, G. (2023). Elephant DOA Estimation using a Geophone Network. In: 2023 26th International Conference on Information Fusion (FUSION): . Paper presented at 26th International Conference on Information Fusion (FUSION), Charleston, USA, 27-30 June 2023. IEEE
Open this publication in new window or tab >>Elephant DOA Estimation using a Geophone Network
Show others...
2023 (English)In: 2023 26th International Conference on Information Fusion (FUSION), IEEE, 2023Conference paper, Published paper (Refereed)
Abstract [en]

Human-wildlife conflicts are a global problem which is central to the Global Goal 15 (life on land). One particular case is elephants, that can cause harm to both people, property and crops. An early warning system that can detect and warn people in time would allow effective mitigation measures. The proposed method is based on a small local network of geophones that sense the seismic waves of elephant footsteps. It is known that elephant footsteps induce low frequency ground waves that can be picked up by geophones in the ground. First, a method is described that detect the particular signature of such footsteps, and then the detections are used to estimate the direction of arrival (DOA). Finally, a Kalman filter is applied to the measurements in order to track the elephant. Field tests performed at a local zoo shows promising results with accurate DOA estimates at 15 meters distance and acceptable accuracy at 40 meters.

Place, publisher, year, edition, pages
IEEE, 2023
Keywords
Meters, Performance evaluation, Location awareness, Seismic measurements, Direction-of-arrival estimation, Target tracking, Prototypes, Elephants, Detection, Direction of Arrival, Kalman filter, Geophone network
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-197793 (URN)10.23919/fusion52260.2023.10224115 (DOI)979-8-89034-485-4 (ISBN)979-8-3503-1320-8 (ISBN)
Conference
26th International Conference on Information Fusion (FUSION), Charleston, USA, 27-30 June 2023
Available from: 2023-09-14 Created: 2023-09-14 Last updated: 2023-09-14
Malmström, M., Skog, I., Axehill, D. & Gustafsson, F. (2023). On the validity of using the delta method for calculating the uncertainty of the predictions from an overparameterized model. In: Hideaki Ishii, Yoshio Ebihara, Jun-ichi Imura, Masaki Yamakita (Ed.), Special issue: 22nd IFAC World Congress: . Paper presented at 22nd IFAC World Congress: Yokohama, Japan, July 9-14, 2023 (pp. 5843-5848). Elsevier, 56(2)
Open this publication in new window or tab >>On the validity of using the delta method for calculating the uncertainty of the predictions from an overparameterized model
2023 (English)In: Special issue: 22nd IFAC World Congress / [ed] Hideaki Ishii, Yoshio Ebihara, Jun-ichi Imura, Masaki Yamakita, Elsevier, 2023, Vol. 56, no 2, p. 5843-5848Conference paper, Published paper (Refereed)
Abstract [en]

The uncertainty in the prediction calculated using the delta method for an over-parameterized (parametric) black-box model is shown to be larger or equal to the uncertainty in the prediction of a canonical (minimal) model. Equality holds if the additional parameters of the overparameterized model do not add flexibility to the model. As a conclusion, for an overparameterized black-box model, the calculated uncertainty in the prediction by the delta method is not underestimated. The results are shown analytically and are validated in a simulation experiment where the relationship between the normalized traction force and the wheel slip of a car is modelled using e.g., a neural network.

Place, publisher, year, edition, pages
Elsevier, 2023
Series
IFAC papersonline, E-ISSN 2405-8963
Keywords
Machine learning; nonlinear system identification; overparameterized model; uncertainty quantification; neural networks; autonomous vehicles
National Category
Control Engineering Communication Systems
Identifiers
urn:nbn:se:liu:diva-199286 (URN)10.1016/j.ifacol.2023.10.077 (DOI)
Conference
22nd IFAC World Congress: Yokohama, Japan, July 9-14, 2023
Available from: 2023-11-24 Created: 2023-11-24 Last updated: 2023-11-24
Forsling, R., Sjanic, Z., Gustafsson, F. & Hendeby, G. (2023). Track-To-Track Association for Fusion of Dimension-Reduced Estimates. In: Proceedings of the 26th International Conference on Information Fusion (FUSION): . Paper presented at 26th International Conference on Information Fusion (FUSION), Charleston, SC, USA, June 27-30, 2023. IEEE
Open this publication in new window or tab >>Track-To-Track Association for Fusion of Dimension-Reduced Estimates
2023 (English)In: Proceedings of the 26th International Conference on Information Fusion (FUSION), IEEE, 2023Conference paper, Published paper (Refereed)
Abstract [en]

Network-centric multitarget tracking under communication constraints is considered, where dimension-reduced track estimates are exchanged. Previous work on target tracking in this subfield has focused on fusion aspects only and derived optimal ways of reducing dimensionality based on fusion performance. In this work we propose a novel problem formalization where estimates are reduced based on association performance. The problem is analyzed theoretically and problem properties are derived. The theoretical analysis leads to an optimization strategy that can be used to partly preserve association quality when reducing the dimensionality of communicated estimates. The applicability of the suggested optimization strategy is demonstrated numerically in a multitarget scenario.

Place, publisher, year, edition, pages
IEEE, 2023
Keywords
Network-centric estimation; Target tracking; Track-to-track association; Communication constraints; Dimension-reduced estimates
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-197014 (URN)10.23919/FUSION52260.2023.10224192 (DOI)979-8-89034-485-4 (ISBN)979-8-3503-1320-8 (ISBN)
Conference
26th International Conference on Information Fusion (FUSION), Charleston, SC, USA, June 27-30, 2023
Funder
Swedish Research Council, Scalable Kalman filtersVinnova, LINK-SICELLIIT - The Linköping‐Lund Initiative on IT and Mobile Communications, 17.12
Note

Funding agency: 10.13039/501100018891-Saab

Available from: 2023-08-17 Created: 2023-08-17 Last updated: 2023-09-03
Zetterqvist, G., Gustafsson, F. & Hendeby, G. (2023). Using Received Power in Microphone Arrays to Estimate Direction of Arrival. In: ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP): . Paper presented at ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Institute of Electrical and Electronics Engineers (IEEE)
Open this publication in new window or tab >>Using Received Power in Microphone Arrays to Estimate Direction of Arrival
2023 (English)In: ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Institute of Electrical and Electronics Engineers (IEEE), 2023Conference paper, Published paper (Refereed)
Abstract [en]

Conventional direction of arrival (DOA) estimators are based on array processing using either time differences or beam-forming. The proposed approach is based on the received power at each microphone, which enables simple hardware, low sampling frequency and small arrays. The problem is recast into a linear regression framework where the least squares method applies, and the main drawback is that different sound sources are not readily separable.Our proposed approach is based on a training phase where the directional sensitivity of each microphone element is estimated. This model is then used as a fingerprint of the observed power vector in a real-time estimator. The learned power vector is here modeled by a Fourier series expansion, which enables Cramér-Rao lower bound computations. We demonstrate the performance using a circular array with eight microphones with promising results.

Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers (IEEE), 2023
Keywords
DOA Estimation, Directional Sensitivity, Microphone Array, CRLB, YALMIP
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-195366 (URN)10.1109/icassp49357.2023.10097197 (DOI)978-1-7281-6327-7 (ISBN)978-1-7281-6328-4 (ISBN)
Conference
ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Note

Funding agencies; G. Zetterqvist has received funding from ELLIIT. This work was partially funded by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation. Theauthors would like to thank Jonas Nordlof at the Swedish Defence Research ¨Agency (FOI) for the help and assistance during the data collection.

Available from: 2023-06-19 Created: 2023-06-19 Last updated: 2023-06-19
Liu, P., Li, K., Hendeby, G. & Gustafsson, F. (2023). Weighted Total Least Squares for Quadratic Errors-in-Variables Regression. In: Proceedings of the 31st Conference on European Signal Processing: . Paper presented at Proceedings of the 31st Conference on European Signal Processing (EUSIPCO), Helsinki, Finland, September 4-8, 2023. (pp. 1894-1897). IEEE
Open this publication in new window or tab >>Weighted Total Least Squares for Quadratic Errors-in-Variables Regression
2023 (English)In: Proceedings of the 31st Conference on European Signal Processing, IEEE, 2023, p. 1894-1897Conference paper, Published paper (Refereed)
Abstract [en]

In this paper, we present a study on using weightedtotal least squares method for parameter estimation of errorsin-variables models with quadratic regressors. The statistics oferror is analyzed to fill in the gap between basic assumptions inweighted total least squares and our case. A modified Cram´er-Rao lower bound is introduced for error quantification in theproposed method. We perform evaluations based on simulationswith comparisons to standard least squares and generalized totalleast squares. Numerical results show that the proposed methodoutperforms the others in terms of estimation accuracy

Place, publisher, year, edition, pages
IEEE, 2023
National Category
Control Engineering Signal Processing
Identifiers
urn:nbn:se:liu:diva-198576 (URN)
Conference
Proceedings of the 31st Conference on European Signal Processing (EUSIPCO), Helsinki, Finland, September 4-8, 2023.
Funder
Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning
Available from: 2023-10-18 Created: 2023-10-18 Last updated: 2023-10-24
Åslund, J., Gustafsson, F. & Hendeby, G. (2023). When Does the Marginalized Particle Filter Degenerate?. In: 2023 26th International Conference on Information Fusion (FUSION): . Paper presented at 26th International Conference on Information Fusion (FUSION), Charleston, South Carolina, USA, June 27-30, 2023. IEEE
Open this publication in new window or tab >>When Does the Marginalized Particle Filter Degenerate?
2023 (English)In: 2023 26th International Conference on Information Fusion (FUSION), IEEE, 2023Conference paper, Published paper (Refereed)
Abstract [en]

The Particle filter can in theory estimate the state of any nonlinear system, but in practice it suffers from an exponential complexity in terms of the number of particles as the dimension of the state increases. The marginalized particle filter can potentially reduce this problem by improving the estimates, particularly for lower number of particles. However, it turns out that for certain systems, it does not provide any improvement in the accuracy of the estimate. The core cause of degeneracy is linked to when the uncertainty of the linear state conditioned on the nonlinear state is 0. Conditions for determining when this occurs are presented and applied to common constant velocity, constant acceleration and constant jerk models with various sampling methods. Interestingly, some combinations are useful while others should be avoided. These findings are supported using simulated systems.

Place, publisher, year, edition, pages
IEEE, 2023
Keywords
Marginalized particle filter, Rao-Blackwellized particle filter, Variance reduction, Particle filter
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-197768 (URN)10.23919/FUSION52260.2023.10224082 (DOI)979-8-3503-1320-8 (ISBN)979-8-89034-485-4 (ISBN)
Conference
26th International Conference on Information Fusion (FUSION), Charleston, South Carolina, USA, June 27-30, 2023
Funder
Swedish Research Council
Available from: 2023-09-13 Created: 2023-09-13 Last updated: 2023-09-14
Forsling, R., Hansson, A., Gustafsson, F., Sjanic, Z., Löfberg, J. & Hendeby, G. (2022). Conservative Linear Unbiased Estimation Under Partially Known Covariances. IEEE Transactions on Signal Processing, 70, 3123-3135
Open this publication in new window or tab >>Conservative Linear Unbiased Estimation Under Partially Known Covariances
Show others...
2022 (English)In: IEEE Transactions on Signal Processing, ISSN 1053-587X, E-ISSN 1941-0476, Vol. 70, p. 3123-3135Article in journal (Refereed) Published
Abstract [en]

Mean square error optimal estimation requires the full correlation structure to be available. Unfortunately, it is not always possible to maintain full knowledge about the correlations. One example is decentralized data fusion where the cross-correlations between estimates are unknown, partly due to information sharing. To avoid underestimating the covariance of an estimate in such situations, conservative estimation is one option. In this paper the conservative linear unbiased estimator is formalized including optimality criteria. Fundamental bounds of the optimal conservative linear unbiased estimator are derived. A main contribution is a general approach for computing the proposed estimator based on robust optimization. Furthermore, it is shown that several existing estimation algorithms are special cases of the optimal conservative linear unbiased estimator. An evaluation verifies the theoretical considerations and shows that the optimization based approach performs better than existing conservative estimation methods in certain cases.

Place, publisher, year, edition, pages
IEEE, 2022
Keywords
Estimation, Optimization, Correlation, Uncertainty, Linear regression, Linear matrix inequalities, Symmetric matrices; Conservative estimation; robust optimization; unknown cross-correlations; covariance intersection; decentralized estimation
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-187807 (URN)10.1109/tsp.2022.3179841 (DOI)000819819300007 ()2-s2.0-85131716447 (Scopus ID)
Note

Funding Agencies|Center for Industrial Information Technology at Linkoping University [17.12]; Industry Excellence Center LINK-SIC - Swedish Governmental Agency for Innovation Systems; Saab AB; Project Scalable Kalman filters - Swedish Research Council

Available from: 2022-08-25 Created: 2022-08-25 Last updated: 2022-09-02Bibliographically approved
Malmström, M., Skog, I., Axehill, D. & Gustafsson, F. (2022). Detection of outliers in classification by using quantified uncertainty in neural networks. In: 25th International Conference of Information Fusion: . Paper presented at 25th International Conference of Information Fusion, FUSION 2022, July 4-7, 2022, Linköping, Sweden. IEEE
Open this publication in new window or tab >>Detection of outliers in classification by using quantified uncertainty in neural networks
2022 (English)In: 25th International Conference of Information Fusion, IEEE, 2022Conference paper, Published paper (Refereed)
Abstract [en]

Neural Networks (NNS) can solve very hard classification and estimation tasks but are less well suited to solve complex sensor fusion challenges, such as end-to-end control of autonomous vehicles. Nevertheless, NN can still be a powerful tool for particular sub-problems in sensor fusion. This would require a reliable and quantifiable measure of the stochastic uncertainty in the predictions that can be compared to classical sensor measurements. However, current NN'S output some figure of merit, that is only a relative model fit and not a stochastic uncertainty. We propose to embed the NN'S in a proper stochastic system identification framework. In the training phase, the stochastic uncertainty of the parameters in the (last layers of the) NN is quantified. We show that this can be done recursively with very few extra computations. In the classification phase, Monte-Carlo (MC) samples are used to generate a set of classifier outputs. From this set, a distribution of the classifier output is obtained, which represents a proper description of the stochastic uncertainty of the predictions. We also show how to use the calculated uncertainty for outlier detection by including an artificial outlier class. In this way, the NN fits a sensor fusion framework much better. We evaluate the approach on images of handwritten digits. The proposed method is shown to be on par with MC dropout, while having lower computational complexity, and the outlier detection almost completely eliminates false classifications.

Place, publisher, year, edition, pages
IEEE, 2022
Keywords
Training, Uncertainty, Current measurement, Stochastic systems, Measurement uncertainty, Stochastic processes, Artificial neural networks
National Category
Control Engineering
Identifiers
urn:nbn:se:liu:diva-188333 (URN)10.23919/FUSION49751.2022.9841376 (DOI)000855689000146 ()9781665489416 (ISBN)9781737749721 (ISBN)
Conference
25th International Conference of Information Fusion, FUSION 2022, July 4-7, 2022, Linköping, Sweden
Note

Funding: Swedens innovation agency, Vinnova, through project iQDeep [2018-02700]

Available from: 2022-09-09 Created: 2022-09-09 Last updated: 2023-10-17
Organisations
Identifiers
ORCID iD: ORCID iD iconorcid.org/0000-0003-3270-171X

Search in DiVA

Show all publications