Jean-Luc Collette

Automatic Characterisation of Musical Signals: Feature Extraction and Temporal Segmentation

This paper presents some results on automatic characterisation of musical and acoustic signals in terms of features attributed to signal segments. These features describe some of the musical and acoustical content of the sound and can be used in applications such as intelligent sound processing, retrieval of music and sound in databases or music editing and labeling. The paper describes research in a very advanced stage but still ongoing. Applications and results on various examples are presented.

Functional semi-automated segmentation of renal DCE-MRI sequences

In dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), segmentation of internal kidney structures is essential for functional evaluation. Manual morphological segmentation of cortex, medulla and cavities remains difficult and time-consuming especially because the different renal compartments are hard to distinguish on a single image. We propose to test a semi-automated method to segment internal kidney structures from a DCE-MRI registered sequence. As the temporal intensity evolution is different in each of the three kidney compartments, pixels are sorted according to their time-intensity curves using a k-means partitioning algorithm. No ground truth is available to evaluate resulting segmentations so a manual segmentation by a radiologist is chosen as a reference. We first evaluate some similarity criteria between the functional segmentations and this reference. The same measures are then computed between another manual segmentation and the reference. Results are similar for the two types of comparisons.

Functional Segmentation of Renal DCE-MRI Sequences Using Vector Quantization Algorithms

In dynamic contrast-enhanced magnetic resonance imaging, segmentation of internal kidney structures like cortex, medulla and cavities is essential for functional assessment. To avoid fastidious and time-consuming manual segmentation, semi-automatic methods have been recently developed. Some of them use the differences between temporal contrast evolution in each anatomical region to perform functional segmentation. We test two methods where pixels are classified according to their time–intensity evolution. They both require a vector quantization stage with some unsupervised learning algorithm (K-means or Growing Neural Gas with targeting). Three or more classes are thus obtained. In the first case the method is completely automatic. In the second case, a restricted intervention by an observer is required for merging. As no ground truth is available for result evaluation, a manual anatomical segmentation is considered as a reference. Some discrepancy criteria like overlap, extra pixels and similarity index are computed between this segmentation and a functional one. The same criteria are also evaluated between the reference and another manual segmentation. Results are comparable for the two types of comparisons, proving that anatomical segmentation can be performed using functional information.

Functional semi-automated segmentation of renal DCE-MRI sequences using a Growing Neural Gas algorithm

In dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), segmentation of internal kidney structures like cortex, medulla and pelvo-caliceal cavities is necessary for functional assessment. Manual segmentation by a radiologist is fairly delicate because images are blurred and highly noisy. Moreover the different compartments cannot be delineated on a single image because they are not visible during the same perfusion phase for physiological reasons. Nevertheless the differences between temporal evolution of contrast in each anatomical region can be used to perform functional segmentation. We propose to test a semi-automated split and merge method based on time-intensity curves of renal pixels. Its first step requires a variant of the classical Growing Neural Gas algorithm. In the absence of ground truth for results assessment, a manual anatomical segmentation by a radiologist is considered as a reference. Some discrepancy criteria are computed between this segmentation and the functional one. As a comparison, the same criteria are evaluated between the reference and another manual segmentation.

On-line restoration for turbulence degraded video in nuclear power plant reactors

This article deals with video inspection of nuclear plant reactor after fuel reloading. During these underwater inspections, the fuel assemblies’ heat generates turbulence effect that sensitively degrades the video quality. An on-line restoration algorithm is proposed here. It consists of two steps. A temporal infinite impulse response filter is first used to get a stabilized but blurry video. A second spatial Wiener deconvolution filter is then used to estimate the video which would have been observed without turbulence. This second filter is based on a probabilistic model of the turbulence impact on the observed video. An on-line prototype, based on this algorithm and its straightforward extension to a moving camera (translation), has been successfully tested on several power plants.

On hybrid filter bank A/D converters with arbitrary over-sampling rate

Hybrid filter banks (HFB) analog/digital (A/D) systems permit wide-band, high frequency conversion. This paper presents a method for designing output digital filters of the HFB, when analog input filters are easy-to-implement (typically second order) and consequently can work at high rate. The constraint of quantization noise amplification due to the digital output filters is taken into account by using the Lagrange multiplier method. In order to improve quality of the output signal in spite of this constraint, degrees of freedom are added by using a K-channels HFB associated with an oversampling factor M less than K and by imposing a condition stronger than that of Nyquist on the input signal.

On Simulations About the Precision of Non Uniform Hybrid Filter Bank Analog/Digital Converters

In this paper we present a theoretical study of the errors due to imperfect reconstruction and to quantization in hybrid filter bank (HFB) analog/digital (A/D) converters that are non-uniform i.e., with K channels and an arbitrary ratio M of the M-fold expanders (M les K). We propose then a new method of simulation for such HFB A/D converters that avoids any numerical computation of differential equation solutions. The results of simulation are compared with theoretical ones. Moreover, in HFB A/D converter studies, the effect of quantization noise is generally indicated by its average power at the output of the HFB, when it is M-cyclostationary. This paper shows on an example that the variance of the global error at the HFB output can vary a lot in a period of M samples

Morphological recognition with the addition of multi-band fluorescence excitation of chlorophylls of phytoplankton

The recognition of aquatic organisms plays a crucial role in the monitoring of the pollution and for the adoption of rapid preventive actions. A compact microscopic optical imaging system is proposed in order to acquire and treat the multibands fluorescence of several pigments in phytoplankton organisms. Two algorithms for automatic recognition of phytoplankton were proposed with a minimum number of calibration parameters. The first algorithm provides a morphological recognition based on “watershed” segmentation and Fourier descriptors, while the second one builds fluorescence pigment images by “k-means” partition of intensity ratios. The operation of these algorithms was illustrated by the study of two different organisms: a cyanobacteria (Dolichospermum sp.) and an alga (Cladophora sp.). The family and the genus of these organisms were then classified into a database which is independent of the size, the orientation and the position of the specimens in the images.

Separation of Impulsive Blended Seismic Sources Using Orthogonal Matching Pursuit

Simultaneous-source (or blended) seismic data acquisition allows reducing acquisition time, which is beneficial in harsh meteorological environment or when strict environmental regulations are applied. The only draw-back of blended acquisition is the interference (or cross-talk) between signals originating from different seismic sources firing at the same time. Recent advances in processing and imaging allow acceptable handling of the cross-talk, however, specific processing methods adapted for blended data still need to be improved. Whether the deblending step (or separation of signals originating from different sources) is necessary remains an open question, but it is still included in the beginning of most of the simultaneous-source processing sequences. In this paper, we propose a deblending method based on the decomposition of the blended signal into a set of locally coherent features, or seismic events. The information on the source contained in each seismic event is further used for separation. The decomposition is performed using the Orthogonal Matching Pursuit signal decomposition algorithm with a specific parametric dictionary adapted for seismic events and allowing sparse representation of the data. The method shows promising results on synthetic sets of 2D seismic data and is scalable to large datasets of industrial size.

Phytoplankton identification by combined methods of morphological processing and fluorescence imaging

The identification of phytoplankton is currently an important issue to prevent the aquatic environment. The growth of one or several phytoplankton species can lead to hyper eutrophication and causes lethal consequences on other organisms. In this paper, the selective recognition of invading species is investigated by automatic recognition algorithms of optical and fluorescence imaging. Firstly, morphological characteristics of algae of microscopic imaging are treated. The image processing lead to the identification the genus of aquatic organisms and compared to a morphologic data base. Secondly, fluorescence images allow an automatic recognition based on multispectral data that identify locally the ratio of different photosynthetic pigments and gives a unique finger print of algae. It is shown that the combination of both methods are useful in the recognition of aquatic organisms.