Algorithm Development

AstroSoft create custom algorithms for software of different applications. We rely on modern computation methods, mathematical and physical base, and most effective and widely used algorithms for solving certain problems.
Quality and efficiency are in prior for us and solution itself can be performed as detailed manual or as complete software.

 

We perform in following areas:

  • 01Sequential analysis for real-time systems (Wald analysis)
    We solve the issuses of mathematical statistics where the number of observations is not fixed in advance, but is fixed based on input data evaluation: either much earlier than in classical hypothesis testing procedures or later if the data appear less informative. Once an observation is complete, a decision is made: either accept the hypothesis under test, discard it, or continue the observation. The use of sequential analysis procedure often considerably reduces decision-making time compared to any optimal method with a fixed number of observations.

    Application fields: data analysis in economics, medicine, ecology and wide range of other data acquisition systems. For example, target detection in electronically agile beam radar is based on Wald analysis.
  • 02Spectral analysis (atomic and molecular)
    Our numerical methods can be used in determining the quantitative and qualitative composition of a substance tested by analyzing the Fraunhofer spectra themselves and their interaction with electromagnetic radiation and acoustic waves, allowing for the distribution of elementary particles by mass and energy. Atomic and molecular composition is determined using emission and absorption methods. Conclusions on the isotopic composition of the substance can be made based on particle mass and energy analysis.

    Application fields: astronomy, metallurgy, criminalistics, medicine, geology, archeology, etc.
  • 03Machine learning
    We create algorithms based on our expertise in mathematical statistics, probability theory, and numerical optimization methods. Algorithms are capable of self-learning by detecting regularities in pre-loaded either collected data. We solve problems of classification, clustering, regression, prediction, reducing data dimension, determining the probability density of data, and finding data dependencies.

    Application fields: automation of the decision-making process in voice and image recognition, classification of documents, technical and medical diagnostics, detecting fraudulent and malicious activities, and much more.
  • 04Signal processing
    Our solutions are used in digital processing of video and radio signals: filtering, enhancement of signal-to-noise ratio, noise suppression, signal discrimination from interference and noise, and correlation analysis. When using digital signal processing (DSP) an evaluation of the speed and efficiency of the proposed processing is typically performed. One-dimensional, two-dimensional or multi-dimensional digital signal processing and quadrature analysis can be used depending on the input signal type and the desired outcome.

    Application fields: radiolocation, acoustics, image processing and data transfer systems. Digital processing procedures applied differ depending on the nature of signal (i.e. signal frequency, modulation, and mode of transmission).

Contact us

If you want more information fill in this form. You will be contacted as soon as possible. Please fill in all required fields.

Get In Touch

  • Gelsingforsskaya 3, corpus 11D
    194044, St. Petersburg, Russia
  • +79522068225
  • This email address is being protected from spambots. You need JavaScript enabled to view it.