Cross-channel Correlation Research Articles

Monaural Speech Segregation Based on Pitch Tracking and Amplitude Modulation

Segregating speech from one monaural recording has proven to be very challenging. Monaural segregation of voiced speech has been studied in previous systems that incorporate auditory scene analysis principles. A major problem for these systems is their inability to deal with the high-frequency part of speech. Psychoacoustic evidence suggests that different perceptual mechanisms are involved in handling resolved and unresolved harmonics. We propose a novel system for voiced speech segregation that segregates resolved and unresolved harmonics differently. For resolved harmonics, the system generates segments based on temporal continuity and cross-channel correlation, and groups them according to their periodicities. For unresolved harmonics, it generates segments based on common amplitude modulation (AM) in addition to temporal continuity and groups them according to AM rates. Underlying the segregation process is a pitch contour that is first estimated from speech segregated according to dominant pitch and then adjusted according to psychoacoustic constraints. Our system is systematically evaluated and compared with pervious systems, and it yields substantially better performance, especially for the high-frequency part of speech.

Color interpolation with variable color ratio considering cross-channel correlation

A color interpolation algorithm is necessary to acquire color images of good quality with a digital still camera in which a single CCD image sensor is equiped. Since a color filter array is adopted, we must interpolate missing color values in the red, green, and blue channels at each pixel location. Inappropriate color interpolation causes color arti- facts, resulting in degraded color resolution. A color interpolation algo- rithm with a variable color ratio is proposed by incorporating cross- channel correlation. The color ratio of weighted sums in the color channels is utilized instead of the weighted sum of color ratios, and the color ratio varies with respect to the cross-channel correlation at each pixel. The proposed algorithm is composed of initial color interpolation step and the updating step of the green channel, where the updating step is carried out iteratively. Furthermore, a minus weighted sum based on the external divide, a modified edge indicator function, and a 5 35 support region for the red and blue channels are also adopted to improve the performance. With the proposed algorithm, we have achieved im- provements in the image quality especially along the edges and in the detail of the image by removing color artifacts. The superiority of the proposed algorithm is demonstrated in the experimental results. © 2004

High definition color interpolation scheme for progressive scan CCD image sensor

A color interpolation scheme for a progressive scan CCD image sensor with a RGB color filter array is required to overcome the physical limitation of the CCD image sensor and to increase the resolution of color signals. Most conventional approaches result in blurred edges and false color artifacts. In this paper, we propose an improved edge-adaptive color interpolation scheme for a progressive scan CCD image sensor. The proposed edge indicator function uses not only the within-channel correlation but also the cross-channel correlation, and reflects the edge characteristics in an image adaptively. The color components unavailable at each channel are interpolated along the edge direction, not across the edges, so that aliasing artifacts are suppressed. Furthermore, we eliminated false color artifacts resulting from the color image formation model in the edge-adaptive color interpolation scheme by adopting the switching algorithm that is based on the color edge detection. Simulation results of the proposed algorithm indicate that the improved edge-adaptive color interpolation scheme produces better quantitative and visually pleasing results than other conventional approaches.

Coherence estimation for SAR imagery

In dual- or multiple-channel synthetic aperture radar (SAR) imaging modes, cross-channel correlation is a potential source of information. The sample coherence magnitude is calculated over a moving window to generate a coherence magnitude map. High-resolution coherence maps may be useful to discriminate fine structures. Coarser resolution is needed for a more accurate estimation of the coherence magnitude. In this study, the accuracy of coherence estimation is investigated as a function of the coherence map resolution. It is shown that the space-averaged coherence magnitude is biased toward higher values. The accuracy of the coherence magnitude estimate obtained is a function of the number of pixels averaged and the number of independent samples per pixel (i.e., the coherence map resolution). A method is proposed to remove the bias from the space-averaged sample coherence magnitude. Coherence magnitude estimation from complex (magnitude and phase) coherence maps is also considered. It is established that the magnitude of the averaged sample coherence estimate is slightly biased for high-resolution coherence maps and that the bias reduces with coarser resolution. Finally, coherence estimation for nonstationary targets is discussed. It is shown that the averaged sample coherence obtained from complex coherence maps or coherence magnitude maps is suitable for estimation of nonstationary coherence. The averaged sample (complex) coherence permits the calculation of an unbiased coherence estimate, provided that the original signals can be assumed to be locally stationary over a sufficiently coarse resolution cell.

An auditory cortical theory of primitive auditory grouping

Auditory stream segregation is considered from the point of view of a theory of the central auditory system in which temporal information is spatially coded on three dimensions roughly corresponding to the level of the cochlea, the inferior colliculus and the auditory cortex, respectively. This theory is embodied in the form of a computational model which simulates peripheral and central processing including a cortical cross-channel correlation mechanism.

A fast converging adaptive filter algorithm for filtered-x systems.

The convergence speed of a multiple input adaptive filtering system is often very slow with the conventional NLMS algorithm. This paper describes a new projection algorithm for filtered-x systems with multiple input signals, secondary sources and control points. This algorithm increase the convergence speed of multiple input adaptive filtering systems by eliminating the cross-channel correlation as well as the auto correlation. However, the computational complexity for the proposed algorithm remains only as much as that of the NLMS algorithm. Computer simulation shows twice faster convergence speed than that of the NLMS algorithm, with only a 1.2 fold increase in computation.

Multichannel signal detection involving temporal and cross-channel correlation

This paper considers the Gaussian multichannel binary detection problem. A multichannel generalized likelihood ratio is derived using a model-based approach where the signal and nonwhite additive noise are characterized by autoregressive (AR) vector processes. Detection performance is obtained via the Monte Carlo procedure for the special case where the AR process parameters of the observation processes are assumed to be known. Specifically, detection performance is obtained for two-channel observation vectors containing various temporal and cross-channel correlation on both the signal and nonwhite additive noise. The computed results are compared with known optimal detection curves using coherent (both in time and across channels) as well as noncoherent integration. In addition, the degradation in performance due to a low signal-to-noise ratio on a specific channel is also considered.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Stochastic modeling and estimation of multispectral image data

Multispectral images consist of multiple channels, each containing data acquired from a different band within the frequency spectrum. Since most objects emit or reflect energy over a large spectral bandwidth, there usually exists a significant correlation between channels. Due to often harsh imaging environments, the acquired data may be degraded by both blur and noise. Simply applying a monochromatic restoration algorithm to each frequency band ignores the cross-channel correlation present within a multispectral image. A Gibbs prior is proposed for multispectral data modeled as a Markov random field, containing both spatial and spectral cliques. Spatial components of the model use a nonlinear operator to preserve discontinuities within each frequency band, while spectral components incorporate nonstationary cross-channel correlations. The multispectral model is used in a Bayesian algorithm for the restoration of color images, in which the resulting nonlinear estimates are shown to be quantitatively and visually superior to linear estimates generated by multichannel Wiener and least squares restoration.

Synthesis of correlated multichannel random processes

This paper describes a method for synthesizing multichannel autoregressive (AR) random processes. The procedure allows for the variation of temporal and cross-channel correlation subject to specific constraints for correlation functions. The resulting synthesized processes provide a "fit" in a minimum mean squared error (MMSE) sense to the process correlation functions specified in terms of their temporal and cross-channel correlation parameters. Computer simulation results are presented showing the case of a two-channel AR process with various values of temporal and cross-channel correlation. A method is also suggested to synthesize a more general class of Gaussian processes with unconstrained quadrature components.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Comment on "Evidence for superdeformed shape isomeric states in 28Si at excitations above 40 MeV through observations of selective particle decays of 16O+12C resonances in 8Be and alpha channels"

A recent study of the $^{12}\mathrm{C}$${(}^{16}$O${,}^{8}$Be${)}^{20}$Ne and $^{12}\mathrm{C}$${(}^{16}$O,\ensuremath{\alpha}${)}^{24}$Mg reactions has led, through cross channel correlation analysis, to the identification of two superdeformed shape isomeric states in $^{28}\mathrm{Si}$ at excitation energies of 43.7 and 46.2 MeV. We suggest that both these states may have \ensuremath{\alpha}-\ensuremath{\alpha}${\mathrm{\ensuremath{-}}}^{16}$O-\ensuremath{\alpha} nuclear molecule structure. Our suggestion is supported theoretically by cranked cluster model calculations and experimentally by previous studies of the $^{12}\mathrm{C}$${(}^{12}$C,\ensuremath{\alpha}\ensuremath{\alpha}${)}^{16}$O reaction, $^{12}\mathrm{C}$${+}^{12}$C scattering, and the $^{12}\mathrm{C}$${(}^{20}$Ne${,}^{12}$${\mathrm{C}}^{12}$C${)}^{8}$Be transfer reaction.

Dynamic formant tracking of noisy speech using temporal analysis on outputs from a nonlinear cochlear model.

In this paper we take a modeling approach to studying representation of formant frequencies of spoken speech and speech in noise in the temporal responses of the peripheral auditory system. On the basis of the properties of the representation, we have devised and evaluated a cross-channel correlation algorithm and an interpeak interval analysis for automatic formant extraction of speech which is strongly dynamic in acoustic characteristics and is embedded in noise. The basilar membrane model used in this study contains laterally coupled damping elements, which are made monotonically dependent on the spatial distribution of the short-term power in the outputs of the model. Efficient digital implementation and the related salient numerical properties of the model are described. Simulation results from the model in response to speech and speech in noise illustrate temporal response patterns that are tonotopically organized in relation to speech formant parameters with little influence by the noise level. By utilizing such relations the devised cross-channel correlation algorithm is shown to be capable of accurately tracking formant movements in spoken syllables and sentences.

Evidence for superdeformed shape isomeric states in 28Si at excitations above 40 MeV through observations of selective particle decays of 16O+12C resonances in 8Be and alpha channels.

Excitation functions of the reactions $^{12}\mathrm{C}$${(}^{16}$O${,}^{8}$Be${)}^{20}$Ne and $^{12}\mathrm{C}$${(}^{16}$O${,}^{4}$He${)}^{24}$Mg leading to several states have been measured in the range ${\mathit{E}}_{\mathrm{c}.\mathrm{m}.}$=25.7--38.6 MeV. From cross channel correlation analysis two intermediate resonance structures at ${\mathit{E}}_{\mathrm{c}.\mathrm{m}.}$=26.9 and 29.5 MeV corresponding to excitation energies of 43.7 and 46.2 MeV in $^{28}\mathrm{Si}$ have been identified. Angular distribution measurements for $^{12}\mathrm{C}$${(}^{16}$O${,}^{4}$He${)}^{24}$${\mathrm{Mg}}_{\mathrm{g}.\mathrm{s}.}$ have also been made at the lower resonance at ${\mathit{E}}_{\mathrm{c}.\mathrm{m}.}$=26.9 MeV which enable spin assignment of ${\mathit{J}}^{\mathrm{\ensuremath{\pi}}}$=${14}^{+}$ to this resonance. Striking observation of preferential $^{8}\mathrm{Be}$ decay of both the resonances to the 8-particle--4-hole states in $^{20}\mathrm{Ne}$ is interpreted as evidence for the resonances to be highly deformed shape isomeric configurations in $^{28}\mathrm{Si}$ expected from deformed shell-model calculations.

A beamformer for the acquisition of evoked potentials

Evoked potentials (EP) contain information about various physiological parameters and the estimation and detection of these signals can aid in the diagnosis of many pathological conditions. However, the signal-to-noise ratio (SNR) for EP measurement is often very low, and thus signal processing techniques must be employed to enhance the SNR. A delay and sum beamformer acquisition system has the potential for significant SNR improvement in EP measurements. In this communication it is shown that an electrode array acquisition system implements a uniform coherent delay and sum beamformer. The performance of the beamformer is characterized in terms of the number of electrodes, and cross-channel correlation. When compared to conventional ensemble averaging, the beamformer reduces the number of response repetitions required to achieve a given SNR by a factor which approaches the number of channels in the acquisition system.

Resonant structures in24Mg+28Si elastic and inelastic scattering

The data on the excitation functions of24Mg+28Si elastic and inelastic (2+ −0+, 2+ −2+, 4+ −0+ and 4+ −2+) scattering fromE c.m.=48.97 to 57.21 MeV have been subjected to a statistical analysis consisting of the calculations of deviation function, cross-correlation function, summed excitation function, cross-channel correlation coefficients, coherence widths, and the distribution of cross sections. Based on the outcome of the analysis resonant structures atE c.m.=49.23, 50.02, 50.51, 52.10, 52.53, 53.27 and 54.14 MeV have been confirmed and three new structures of the same nature have been identified atE c.m.=51.42, 54.88 and 55.60 MeV.

Nonstatistical structures in the excitation functions of20Ne(16O,12C)24Mg and20Ne(16O,20Ne)16O reactions

The data on the excitation functions of20Ne(16O,12C)24Mg,20Ne(16O,12C)24Mg*(1.37, 2+),20Ne(16O,12C)24Mg*(4.12, 4++4.24, 2+) +20Ne(16O,12C*(4.44, 2+))24Mg,20Ne(16O,12C)24Mg*(6.01, 4++6.43, 0+),20Ne(16O,20Ne)16O,20Ne(16O,20Ne*(1.63, 2+))16O, and20Ne(16O,20Ne*(4.25, 4+))16O reactions atθlab=13° fromEc.m.=22.8 to 38.6 MeV have been subjected to a statistical analysis comprising of the calculations of the distribution of cross sections, deviation functions, cross-correlation functions, summed excitation functions, cross-channel correlation coefficients and coherence widths. The analysis confirms the existence of nonstatistical structures atEc.m.=24.6, 27.8, 31.7 and 35.5 MeV, and identifies a new structure of the same nature atEc.m. =25.6 MeV.

Statistical analysis of24Mg+24Mg elastic and inelastic scattering

The data on the excitation functions of24Mg+24Mg elastic and inelastic (24Mg +24Mg*(2+),24Mg*(2+)+24Mg*(2+),24Mg+24Mg*(4+),24Mg*(4+)+24Mg*(2+),24Mg+24Mg*(6+)) scattering fromE c.m=42 to 56 MeV have been subjected to a statistical analysis consisting of calculations of deviation function, cross-correlation function, cross-channel correlation coefficients, coherence widths, and the distribution of cross sections. On the basis of the analysis resonant structures atE c.m=45.70, 46.65, 47.35 and 47.75 MeV have been confirmed. Two new resonant structures atE c.m=44.55 and 50.50 MeV have been identified.

A composite model of the auditory periphery for the processing of speech

A composite model of the auditory periphery has been developed to process speech sounds in a manner consistent with physiological responses recorded from the mammalian auditory nerve. The model consists of discrete stages simulating (1) the middle ear, (2) the motion of the basilar membrane, (3) the receptor potential of inner hair cells and (4) the synaptic properties of the inner-hair-cell/auditory-nerve-fiber junction. The spatio-temporal discharge pattern of the auditory nerve is displayed as the output of the composite model. It is shown that the output of the model, responding in a nonlinear manner, most accurately simulates the coherent synchrony patterns of auditory-nerve fibers distributed over a broad range of characteristic frequencies evoked by speech sounds. It is further shown that such a nonlinear model is capable of preserving the basic spatio-temporal representation of speech under low signal-to-noise conditions. A cross-channel correlation technique can be used to extract formant information from the distribution of temporal discharge activity, providing a spectral representation of the signal which is relatively invariant over a wide range of signal-to-noise ratios.

A composite auditory model for processing speech sounds.

A composite inner-ear model, containing the middle ear, basilar membrane (BM), hair cells, and hair-cell/nerve-fiber synapses, is presented. The model incorporates either a linear-BM stage or a nonlinear one. The model with the nonlinear BM generally shows a high degree of success in reproducing the qualitative aspects of experimentally recorded cat auditory-nerve-fiber responses to speech. In modeling fiber population responses to speech and speech in noise, it was found that the BM nonlinearity allows bands of fibers in the model to synchronize strongly to a common spectral peak in the stimulus. A cross-channel correlation algorithm has been devised to further process the model's population outputs. With output from the nonlinear-BM model, the cross-channel correlation values are appreciably reduced only at those channels whose CFs coincide with the formant frequencies. This observation also holds, to a large extent, for noisy speech.

Gamma-ray studies of theC12+C12system

Excitation functions for the yields of ten residual nuclides from the $^{12}\mathrm{C}$+$^{12}\mathrm{C}$ reaction have been measured over the range ${E}_{\mathrm{c}.\mathrm{m}.}=5.25\ensuremath{-}20.0$ MeV in steps of 125 keV, using $\ensuremath{\gamma}$-ray techniques. Nearly all of the reaction channels, including those with light-particle evaporation, showed strong narrow structures. A qualitative statistical analysis performed on the data gave useful information about the locations of possible nonstatistical structure, based primarily on the very strong cross-channel correlation observed in the data set.NUCLEAR REACTIONS $^{12}\mathrm{C}$ ($^{12}\mathrm{C}$, $x$); ${E}_{\mathrm{c}.\mathrm{m}.}=5.25\ensuremath{-}20.0$ MeV; measured $\ensuremath{\sigma}(E)$ for production of $A=16\ensuremath{-}23$ reaction products; statistical analysis presented; deduced evidence for nonstatistical structure.