Gaussian mixture models as automated particle classifiers for fast neutron detectors

Abstract

Pulse shape discrimination (PSD) is the task of classifying electronic pulse shapes for different particle types such as gamma rays and fast neutrons interacting in scintillators and read out by photo sensitive detectors. This field has been limited in its adoption of techniques found in the statistical learning community. Methods initially employed in the 1960s for analog electronic circuitry persist in the current PSD literature describing operations performed on digitized pulses, which are amenable to statistical rigor. Despite vast amounts of data collected at low energy levels, traditional PSD methods are unable to discriminate particles below a certain threshold.In this work, Gaussian mixture models (GMMs) are used as a clustering technique for fast neutron detection in the absence of labeled data. GMMs yield improvements spanning the energy spectrum in a desirably efficient, unsupervised fashion. An extension, the Dirichlet Process GMM, provides further flexibility and classification improvement. K E Y W O R D Sclassification, clustering, mixture models, pulse shape discrimination

Pulse pileup rejection methods using a two-component Gaussian Mixture Model for fast neutron detection with pulse shape discriminating scintillator

“…For the first case, solutions used by the authors rely on trigger algorithms implemented in commercially-available digital pulse processing hardware and firmware [6]. The authors covered the second case in a recent series of papers [7,8,9,10,11]. This paper discusses the third case: triggercollected pulses which exhibit shapes unlike those from either neutrons or gammas.…”

“…In previous papers [7,8,10], working under a practical constraint to use only unsupervised methods, the authors showed that a two-component GMM is a straightforward and effective method of relying on unlabeled training data consisting of neutrons and gammas to obtain a score to discriminate neutrons from gammas. GMM finds clusters in the multivariate space of observed features of the training data that contains both neutrons and gammas.…”

“…The authors selected a specific method of pre-processing the feature vectors that yielded the best neutron-gamma discrimination results for the datacollection setup described in Section 4.1 when scored using labeled data and evaluated with a ROC curve [7]. The user is also free to choose the most appropriate covariance structure for the GMM, and the authors selected a shared covariance matrix (same covariance matrix across the two components), and used only the diagonal entries (variances) to compute the Mahalanobis distance [8], as follows:…”

A Simplified Correlation Index for Fast Real-Time Pulse Shape Recognition

“…Blair et al [ 13 ] used a normalized cross-correlation method to distinguish between two types of pulse shapes. Their algorithm is also based on the Pearson’s correlation.…”

Neutron time correlations with neutron and gamma sensitive scintillators

“…Such scintillators also detect γ-rays and the neutrons need to be separated from the γ-rays. Usual methods involve pulse-shape discrimination (PSD) although newer, more advanced methods involving a Gaussian mixture model (GMM) have also been developed [8,9,10,11,12,13,14].…”