Theory inspired deep network for instantaneous-frequency extraction and signal components recovery from discrete blind-source data
Authors:
Charles K. Chui,
Ningning Han,
Hrushikesh N. Mhaskar
Abstract:
This paper is concerned with the inverse problem of recovering the unknown signal components, along with extraction of their instantaneous frequencies (IFs), governed by the adaptive harmonic model (AHM), from discrete (and possibly non-uniform) samples of the blind-source composite signal.
None of the existing decomposition methods and algorithms, including the most popular empirical mode decom…
▽ More
This paper is concerned with the inverse problem of recovering the unknown signal components, along with extraction of their instantaneous frequencies (IFs), governed by the adaptive harmonic model (AHM), from discrete (and possibly non-uniform) samples of the blind-source composite signal.
None of the existing decomposition methods and algorithms, including the most popular empirical mode decomposition (EMD) computational scheme and its current modifications, is capable of solving this inverse problem.
In order to meet the AHM formulation and to extract the IFs of the decomposed components, called intrinsic mode functions (IMFs), each IMF of EMD is extended to an analytic function in the upper half of the complex plane via the Hilbert transform, followed by taking the real part of the polar form of the analytic extension.
Unfortunately, this approach most often fails to resolve the inverse problem satisfactorily.
More recently, to resolve the inverse problem, the notion of synchrosqueezed wavelet transform (SST) was proposed by Daubechies and Maes, and further developed in many other papers, while a more direct method, called signal separation operation (SSO), was proposed and developed in our previous work published in the journal, Applied and Computational Harmonic Analysis, vol. 30(2):243-261, 2016.
In the present paper, we propose a synthesis of SSO using a deep neural network, based directly on a discrete sample set, that may be non-uniformly sampled, of the blind-source signal.
Our method is localized, as illustrated by a number of numerical examples, including components with different signal arrival and departure times.
It also yields short-term prediction of the signal components, along with their IFs.
Our neural networks are inspired by theory, designed so that they do not require any training in the traditional sense.
△ Less
Submitted 31 January, 2020;
originally announced January 2020.
Bayesian Modeling and Computation for Analyte Quantification in Complex Mixtures Using Raman Spectroscopy
Authors:
Ningren Han,
Rajeev J. Ram
Abstract:
In this work, we propose a two-stage algorithm based on Bayesian modeling and computation aiming at quantifying analyte concentrations or quantities in complex mixtures with Raman spectroscopy. A hierarchical Bayesian model is built for spectral signal analysis, and reversible-jump Markov chain Monte Carlo (RJMCMC) computation is carried out for model selection and spectral variable estimation. Pr…
▽ More
In this work, we propose a two-stage algorithm based on Bayesian modeling and computation aiming at quantifying analyte concentrations or quantities in complex mixtures with Raman spectroscopy. A hierarchical Bayesian model is built for spectral signal analysis, and reversible-jump Markov chain Monte Carlo (RJMCMC) computation is carried out for model selection and spectral variable estimation. Processing is done in two stages. In the first stage, the peak representations for a target analyte spectrum are learned. In the second, the peak variables learned from the first stage are used to estimate the concentration or quantity of the target analyte in a mixture. Numerical experiments validated its quantification performance over a wide range of simulation conditions and established its advantages for analyte quantification tasks under the small training sample size regime over conventional multivariate regression algorithms. We also used our algorithm to analyze experimental spontaneous Raman spectroscopy data collected for glucose concentration estimation in biopharmaceutical process monitoring applications. Our work shows that this algorithm can be a promising complementary tool alongside conventional multivariate regression algorithms in Raman spectroscopy-based mixture quantification studies, especially when collecting a large training dataset with high quality is challenging or resource-intensive.
△ Less
Submitted 19 May, 2018;
originally announced May 2018.
