Jean-Michel Papy

Exponential data fitting using multilinear algebra: the single‐channel and multi‐channel case

There is a wide variety of signal processing applications in which the data are assumed to be modelled as a sum of exponentially damped sinusoids. Many subspace-based approaches (such as ESPRIT, matrix pencil, Prony, etc.) aim to estimate the parameters of this model. Typically, the data are arranged in Toeplitz or Hankel matrices and suitable parameter estimates are obtained via a truncated singular value decomposition (SVD) of the data matrix. It is shown that the parameter accuracy may be improved by arranging single-channel or multi-channel data in a higher-order tensor and estimating the model parameters via a multilinear generalization of the SVD. The algorithm is presented and its performance is illustrated by means of simulations. Copyright ? 2005 John Wiley & Sons, Ltd.

Common pole estimation in multi-channel exponential data modeling

In this paper we develop a technique for detecting and retrieving the common harmonics in a multi-channel setup. The different (complex) signals are arranged in a set of Hankel matrices. First, we compute the singular value decomposition (SVD) of these matrices in order to denoise and normalize the relevant subspaces. Then a second SVD is applied to detect and select the common subspace. The poles of the harmonics are computed from the common subspace using the total least squares (TLS) technique. This method is very flexible, robust and outperforms existing subspace-based methods. Moreover, it can be applied to an arbitrary number of channels.

Modeling common dynamics in multichannel signals with applications to artifact and background removal in EEG recordings

Removing artifacts and background electroencephalography (EEG) from multichannel interictal and ictal EEG has become a major research topic in EEG signal processing in recent years. We applied for this purpose a recently developed subspace-based method for modeling the common dynamics in multichannel signals. When the epileptiform activity is common in the majority of channels and the artifacts appear only in a few channels the proposed method can be used to remove the latter. The performance of the method was tested on simulated data for different noise levels. For high noise levels the method was still able to identify the common dynamics. In addition, the method was applied to real life EEG recordings containing interictal and ictal activity contaminated with muscle artifact. The muscle artifacts were removed successfully. For both the synthetic data and the analyzed real life data the results were compared with the results obtained with principal component analysis (PCA). In both cases, the proposed method performed better than PCA.

Kalman-Filtering-Based Prognostics for Automatic Transmission Clutches

Demands of low-cost prognostics tool for automatic transmission clutches (i.e., based on measurement data from sensors typically available) by industry have increased since the last few years. In this paper, a prognostics tool is developed by fusing a newly developed degradation model with the measurable pre-lockup feature under the extended Kalman filtering framework. As this feature can be extracted from sensory data typically available in wet clutch applications, the developed prognostics tool, hence, does not require extra cost for any additional sensor. New history data of commercially available wet clutches obtained from accelerated life tests using a fully instrumented SAE#2 test setup have been acquired and processed. The experimental results show that the prognostics algorithm developed in this paper outperforms the early developed prognostics algorithm, which is based on the weighted mean slope method (i.e., data-driven approach). It is shown that the clutch remaining useful life estimations with the novel prognostics algorithm remain in the desired accuracy region of 20% with relatively small uncertainty interval in comparison with the early developed prognostics algorithm.

Fibre optic sensor for continuous health monitoring in CFRP composite materials

An intensity modulated sensor, based on the microbending concept, has been incorporated in laminates produced from a C/epoxy prepreg. Pencil lead break tests (Hsu-Neilsen sources) and tensile tests have been performed on this material. In this research study, fibre optic sensors will be proven to offer an alternative for the robust piezoelectric transducers used for Acoustic Emission (AE) monitoring. The main emphasis has been put on the use of advanced signal processing techniques based on time-frequency analysis. The signal Short Time Fourier Transform (STFT) has been computed and several robust noise reduction algorithms, such as Wiener adaptive filtering, improved spectral subtraction filtering, and Singular Value Decomposition (SVD) -based filtering, have been applied. An energy and frequency -based detection criterion is put forward to detect transient signals that can be correlated with Modal Acoustic Emission (MAE) results and thus damage in the composite material. There is a strong indication that time-frequency analysis and the Hankel Total Least Squares (HTLS) method can also be used for damage characterization. This study shows that the signal from a quite simple microbend optical sensor contains information on the elastic energy released whenever damage is being introduced in the host material by mechanical loading. Robust algorithms can be used to retrieve and analyze this information.

Removing Artifacts and Background Activity in Multichannel Electroencephalograms by Enhancing Common Activity

Removing artifacts and background EEG from multichannel interictal and ictal EEG has become a major research topic in EEG signal processing in recent years. We applied for this purpose a recently developed subspace-based method for modelling the common dynamics in multichannel signals. When the epileptiform activity is common in the majority of channels and the artifacts appear only in a few channels the proposed method can be used to remove the latter. The performance of the method was tested on simulated data for different noise levels. For high noise levels the method was still able to identify the common dynamics. In addition, the method was applied to a real life EEG recording. Also in this case the muscle artifacts were removed successfully. For both the synthetic data and the analyzed real life data the results were compared with the results obtained with principal component analysis (PCA). In both cases the proposed method performed better than PCA

Acoustic emission monitoring using a multimode optical fiber sensor

Permanent damage in various materials and constructions often causes high-energy high-frequency acoustic waves. To detect those so called ‘acoustic emission (AE) events’, in most cases ultrasonic transducers are embedded in the structure or attached to its surface. However, for many applications where event localization is less important, an embedded low-cost multimode optical fiber sensor configured for event counting may be a better alternative due to its corrosion resistance, immunity to electromagnetic interference and light-weight. The sensing part of this intensity-modulated sensor consists of a multimode optical fiber. The sensing principle now relies on refractive index variations, microbending and mode-mode interferences by the action of the acoustic pressure wave. A photodiode is used to monitor the intensity of the optical signal and transient signal detection techniques (filtering, frame-to-frame analysis, recursive noise estimation, power detector estimator) on the photodiode output are applied to detect the events. In this work, the acoustic emission monitoring capabilities of the multimode optical fiber sensor are demonstrated with the fiber sensor embedded in the liner of a Power Data Transmission (PDT) coil to detect damage (delamination, matrix cracking and fiber breaking) while bending the coil. With the Hankel Total Least Square (HTLS) technique, it is shown that both the acoustic emission signal and optical signal can be modeled with a sum of exponentially damped complex sinusoids with common poles.

Incremental Classifier Fusion and Its Applications in Industrial Monitoring and Diagnostics

Pattern recognition techniques have shown their usefulness for monitoring and diagnosing many industrial applications. The increasing production rates and the growing databases generated by these applications require learning techniques that can adapt their models incrementally, without revisiting previously used data. Ensembles of classifiers have been shown to improve the predictive accuracy as well as the robustness of classification systems. In this work, several well-known classifier fusion methods (Fuzzy Integral, Decision Templates, Dempster–Shafer Combination, and Discounted Dempster–Shafer Combination) are extended to allow incremental adaptation. Additionally, an incremental classifier fusion method using an evolving clustering approach is introduced—named Incremental Direct Cluster-based ensemble. A framework for strict incremental learning is proposed in which the ensemble and its member classifiers are adapted concurrently. The proposed incremental classifier fusion methods are evaluated within this framework for two industrial applications: online visual quality inspection of CD imprints and prediction of maintenance actions for copiers from a large historical database.