kSpectra is a scientific application for advanced time series analysis. It provides a set of powerful tools for analysis of univariate or multivariate time series in various sciences, ranging from electrical engineering and physics to geophysics, as well as life, biomedical sciences, finance and economics. kSpectra helps to identify and predict periodic signals in noisy and short and gappy time series, when standard methods (FFT/wavelets etc) have limited applicability. kSpectra contains procedures for:
- estimating the spectrum, cross-spectrum and coherence,
- decomposing the time series into trends, oscillatory components, and noise by using sophisticated statistical significance tests,
- reconstructing and predicting the contributions of trends and near-periodic components of the time series,
- gap-filling technique for analysis of datasets with missing data.
kSpectra includes Blackman-Tukey Correlogram (BT-FFT), Maximum-Entropy Method (MEM), MultiTaper Method (MTM), Singular Spectrum Analysis (SSA), multichannel SSA (MSSA), and Principal Component Analysis (PCA). The basic philosophy of kSpectra is that only the simultaneous and flexible application of more than one spectral estimation method can provide truly reliable information on a given time series, when the signal-to-noise ratio is low.
The results of kSpectra are mainly plots of spectrograms (power vs. frequencies) and significance tests against noise level from various spectral estimation methods (SSA, MTM, MEM, BT- FFT). The user typically needs to import the input data by using Data I/O panel into named data objects (Vectors or Matrices), and then apply different spectral estimation tools; each tool has it’s own panel of GUI. The user’s task is to interpret obtained spectrograms for presence of oscillatory modes. The spectrograms can be saved in PDF or EPS format.
kSpectra supports data I/O in ASCII, CSV and Matlab formats.
It is a project-based application, and allows to save all the data, results and plots in a singe project file.
