Grant: Discovery Grants Program - Individual
Trent University
$28,000.00
Spectral Methods for Time Series Imputation and Signal Processing
EF 2024-25
Résumé
One fundamental goal for time series analysis is the estimation of temporal structure in scientific data, and the need for accurate and computationally efficient methods continues to grow. Processes from the natural world contain significant periodic structure, and spectral analysis is the appropriate domain for estimation and analysis of such structure. Two areas in which important practical issues remain are: the reconstruction and validation of scientific data sets; and signal detection problems. Time series data sets have value coupled with the availability of contiguous samples, so missing observations are a concern for the validity and use of such data; imputation can extend the length of a data set for increased utility. Scientific time series data contain numerous periodic components ('signals'), with their detection being of high interest. Extending well-understood signal detection algorithms to further classes of data, e.g., nonstationary elements such as frequency-modulated signals or uniformly modulated processes, will expand the ability of analysts to detect and qualify signals in such series.
The proposed research will focus on the development of statistical procedures to meet the challenges posed by signal detection in the presence of noise, and by the need for long time series data sets for the answering of scientific problems. Research challenges arise when trying to estimate temporal structure in the presence of missingness, and development of robust imputation methods suitable for reconstruction of contiguous records (preserving spectral structure) are valuable for increasing the utility of many already-existing time series data. Signal detection algorithms are typically structured with quite firmly defined assumptions, with violations resulting in either poor performance. Expanding current frameworks through broadened assumptions, such as through uniformly modulated processes or frequency-modulated signals, allows for a broader class of possible detections.
The research objectives are divided into two main themes: (1) tools to estimate, calibrate and test imputations of missing observations in time series data sets, including the use of neural network models; and (2) the incorporation of novel spectral methods into existing signal processing techniques; the increase in robustness of frequency domain methods; and the extension to frequency-modulated periodic structure and certain classes of nonstationarity for previously mentioned methods.
The training component of the proposed research will immerse students in current topics in time series analysis connected to the natural sciences. The research is expected to provide tools to increase the usefulness of time series signal processing techniques, to develop cutting-edge imputation methods for time series modeling, and to continue development of statistical signal processing tools toward inclusion of select classes of nonstationarity.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2024-04420
Name nom
Burr, Wesley
Department département
Mathematics
Organization id
32
Institution établissement
Trent University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
28000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1200
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Advancement of knowledge
Area of application fr
Avancement des connaissances
Research subject code
3002
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Time series analysis
Research subject fr
Analyse des séries chronologiques
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Spectral Methods for Time Series Imputation and Signal Processing
Keywords
spectrum estimation; time series; imputation; interpolation; signal processing; signal detection; environmental data
Application summary
One fundamental goal for time series analysis is the estimation of temporal structure in scientific data, and the need for accurate and computationally efficient methods continues to grow. Processes from the natural world contain significant periodic structure, and spectral analysis is the appropriate domain for estimation and analysis of such structure. Two areas in which important practical issues remain are: the reconstruction and validation of scientific data sets; and signal detection problems. Time series data sets have value coupled with the availability of contiguous samples, so missing observations are a concern for the validity and use of such data; imputation can extend the length of a data set for increased utility. Scientific time series data contain numerous periodic components ('signals'), with their detection being of high interest. Extending well-understood signal detection algorithms to further classes of data, e.g., nonstationary elements such as frequency-modulated signals or uniformly modulated processes, will expand the ability of analysts to detect and qualify signals in such series.
The proposed research will focus on the development of statistical procedures to meet the challenges posed by signal detection in the presence of noise, and by the need for long time series data sets for the answering of scientific problems. Research challenges arise when trying to estimate temporal structure in the presence of missingness, and development of robust imputation methods suitable for reconstruction of contiguous records (preserving spectral structure) are valuable for increasing the utility of many already-existing time series data. Signal detection algorithms are typically structured with quite firmly defined assumptions, with violations resulting in either poor performance. Expanding current frameworks through broadened assumptions, such as through uniformly modulated processes or frequency-modulated signals, allows for a broader class of possible detections.
The research objectives are divided into two main themes: (1) tools to estimate, calibrate and test imputations of missing observations in time series data sets, including the use of neural network models; and (2) the incorporation of novel spectral methods into existing signal processing techniques; the increase in robustness of frequency domain methods; and the extension to frequency-modulated periodic structure and certain classes of nonstationarity for previously mentioned methods.
The training component of the proposed research will immerse students in current topics in time series analysis connected to the natural sciences. The research is expected to provide tools to increase the usefulness of time series signal processing techniques, to develop cutting-edge imputation methods for time series modeling, and to continue development of statistical signal processing tools toward inclusion of select classes of nonstationarity.