Iterative Non-Negative Matrix Factorization Filter for Blind Deconvolution in Photon/Ion Counting.

Area of Science:

• Analytical Chemistry
• Signal Processing
• Data Science

Background:

• Digital signals from analytical instruments are often convolved with instrument impulse response functions (IRFs).
• This convolution complicates quantitative analysis and data interpretation, especially in techniques like scintillation counting with limited dynamic range.
• Existing deconvolution methods typically require prior knowledge of the IRF, which is difficult to obtain.

Purpose of the Study:

• To develop a blind deconvolution method for temporal data analysis.
• To simultaneously recover photon arrival times and the instrument impulse response function (IRF) from large datasets.
• To apply the developed method to analyze triboluminescence (TL) data for discriminating signal origins.

Main Methods:

• A digital filter based on non-negative matrix factorization (NMF) was employed.
• A sliding window approach was used to perform NMF on short segments of large datasets.
• Random initial guesses were used for the IRF to enable simultaneous recovery of deconvolved photon arrival times and the IRF.

Main Results:

• The NMF filter successfully performed blind deconvolution of temporal information.
• Both deconvolved photon arrival times and the IRF were simultaneously recovered.
• The NMF filter enabled discrimination between different hypothesized physical origins of triboluminescence signals from active pharmaceutical ingredients.

Conclusions:

• Non-negative matrix factorization provides an effective approach for blind deconvolution of temporal data in analytical instrumentation.
• This method overcomes limitations of traditional deconvolution techniques by not requiring prior knowledge of the IRF.
• The NMF filter demonstrates utility in analyzing complex signals, such as those from triboluminescence, for advanced material characterization.