State-of-the-art models for semantic segmentation are based on adaptations of convolutional networks that had originally been designed for image classification. However, dense prediction and image classification are structurally different. In this work, we develop a new convolutional network module that is specifically designed for dense prediction. The presented module uses dilated convolutions to systematically aggregate multi-scale contextual information without losing resolution. The architecture is based on the fact that dilated convolutions support exponential expansion of the receptive field without loss of resolution or coverage. We show that the presented context module increases the accuracy of state-of-the-art semantic segmentation systems. In addition, we examine the adaptation of image classification networks to dense prediction and show that simplifying the adapted network can increase accuracy.

Multi-Scale Context Aggregation by Dilated Convolutions

This article presents, for the first time, a practical method for the direct quantification of frequency-specific synchronization (i.e., transient phase-locking) between two neuroelectric signals. The motivation for its development is to be able to examine the role of neural synchronies as a putative mechanism for long-range neural integration during cognitive tasks. The method, called phase-locking statistics (PLS), measures the significance of the phase covariance between two signals with a reasonable time-resolution (,100 ms). Unlike the more traditional method of spectral coherence, PLS separates the phase and amplitude components and can be directly interpreted in the framework of neural integration. To validate synchrony values against background fluctuations, PLS uses surrogate data and thus makes no a priori assumptions on the nature of the experimental data. We also apply PLS to investigate intracortical recordings from an epileptic patient performing a visual discrimination task. We find large-scale synchronies in the gamma band (45 Hz), e.g., between hippocampus and frontal gyrus, and local synchronies, within a limbic region, a few cm apart. We argue that whereas long-scale effects do reflect cognitive processing, short-scale synchronies are likely to be due to volume conduction. We discuss ways to separate such conduction effects from true signal synchrony. Hum Brain Mapping 8:194-208, 1999. r 1999 Wiley-Liss, Inc.

/pdf/measuring-phase-synchrony-in-brain-signals-wbvxlb4wks.pdf

Measuring phase synchrony in brain signals

Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in their building modules. In this work, we introduce two new modules to enhance the transformation modeling capability of CNNs, namely, deformable convolution and deformable RoI pooling. Both are based on the idea of augmenting the spatial sampling locations in the modules with additional offsets and learning the offsets from the target tasks, without additional supervision. The new modules can readily replace their plain counterparts in existing CNNs and can be easily trained end-to-end by standard back-propagation, giving rise to deformable convolutional networks. Extensive experiments validate the performance of our approach. For the first time, we show that learning dense spatial transformation in deep CNNs is effective for sophisticated vision tasks such as object detection and semantic segmentation. The code is released at https://github.com/msracver/Deformable-ConvNets.

Deformable Convolutional Networks

This paper introduces WaveNet, a deep neural network for generating raw audio
waveforms. The model is fully probabilistic and autoregressive, with the
predictive distribution for each audio sample conditioned on all previous ones;
nonetheless we show that it can be efficiently trained on data with tens of
thousands of samples per second of audio. When applied to text-to-speech, it
yields state-of-the-art performance, with human listeners rating it as
significantly more natural sounding than the best parametric and concatenative
systems for both English and Mandarin. A single WaveNet can capture the
characteristics of many different speakers with equal fidelity, and can switch
between them by conditioning on the speaker identity. When trained to model
music, we find that it generates novel and often highly realistic musical
fragments. We also show that it can be employed as a discriminative model,
returning promising results for phoneme recognition.

WaveNet: A Generative Model for Raw Audio

A simple, nonrigorous, synthetic view of wavelet theory is presented for both review and tutorial purposes. The discussion includes nonstationary signal analysis, scale versus frequency, wavelet analysis and synthesis, scalograms, wavelet frames and orthonormal bases, the discrete-time case, and applications of wavelets in signal processing. The main definitions and properties of wavelet transforms are covered, and connections among the various fields where results have been developed are shown. >

/pdf/wavelets-and-signal-processing-1gyx22a8e3.pdf

Wavelets and signal processing

Wavelets: Time-Frequency Methods and Phase Space

Wavelets. Time-Frequency Methods and Phase Space.

I Introduction to Wavelet Transforms.- Reading and Understanding Continuous Wavelet Transforms.- Orthonormal Wavelets.- Orthonormal Bases of Wavelets with Finite Support - Connection with Discrete Filters.- II Some Topics in Signal Analysis.- Some Aspects of Non-Stationary Signal Processing with Emphasis on Time-Frequency and Time-Scale Methods.- Detection of Abrupt Changes in Signal Processing.- The Computer, Music, and Sound Models.- III Wavelets and Signal Processing.- Wavelets and Seismic Interpretation.- Wavelet Transformations in Signal Detection.- Use of Wavelet Transforms in the Study of Propagation of Transient Acoustic Signals Across a Plane Interface Between Two Homogeneous Media.- Time-Frequency Analysis of Signals Related to Scattering Problems in Acoustics Part I: Wigner-Ville Analysis of Echoes Scattered by a Spherical Shell.- Coherence and Projectors in Acoustics.- Wavelets and Granular Analysis of Speech.- Time-Frequency Representations of Broad-Band Signals.- Operator Groups and Ambiguity Functions in Signal Processing.- IV Mathematics and Mathematical Physics.- Wavelet Transform Analysis of Invariant Measures of Some Dynamical Systems.- Holomorphic Integral Representations for the Solutions of the Helmholtz Equation.- Wavelets and Path Integrals.- Mean Value Theorems and Concentration Operators in Bargmann and Bergman Space.- Besov-Sobolev Algebras of Symbols.- Poincare Coherent States and Relativistic Phase Space Analysis.- A Relativistic Wigner Function Affiliated with the Weyl-Poincare Group.- Wavelet Transforms Associated to the n-Dimensional Euclidean Group with Dilations: Signal in More Than One Dimension.- Construction of Wavelets on Open Sets.- Wavelets on Chord-Arc Curves.- Multiresolution Analysis in Non-Homogeneous Media.- About Wavelets and Elliptic Operators.- Towards a Method for Solving Partial Differential Equations Using Wavelet Bases.- V Implementations.- A Real-Time Algorithm for Signal Analysis with the Help of the Wavelet Transform.- An Implementation of the "algorithme a trous" to Compute the Wavelet Transform.- An Algorithm for Fast Imaging of Wavelet Transforms.- Multiresolution Approach to Wavelets in Computer Vision.- Index of Contributors.

Wavelets : time-frequency methods and phase space : proceedings of the international conference, Marseille, France, December 14-18, 1987

A recently introduced algorithm [Frick et al., Astrophys. J. 483, 426 (1997)] of spectral analysis of data with gaps via a modified continuous wavelet transform is developed and studied. This algorithm is based on a family of functions called "gapped wavelets" which fulfill the admissibility condition on the gapped support. The wavelet family is characterized by an additional parameter which should be calculated for every scale and position. Three theorems concerning the properties of gapped wavelet transform are formulated and proved. They affirm the global stability of the algorithm as well as its stability in both limits of large and small scales. These results are illustrated by some numerical examples, which show that the algorithm really attenuates the artifacts coming from gaps (and/or boundaries), and is particularly efficient at small and large scales.

Wavelet analysis of signals with gaps

The Schrodinger scattering amplitude for a fixed potential is studied as a function of the three components of the initial momentum, the three components of the final momentum, and the square root of the energy.

Alexander Grossmann

Papers

Wavelets: Time-Frequency Methods and Phase Space

Wavelets. Time-Frequency Methods and Phase Space.

Wavelets : time-frequency methods and phase space : proceedings of the international conference, Marseille, France, December 14-18, 1987

Wavelet analysis of signals with gaps

Schrodinger Scattering Amplitude. III