Abstracts of 3rd SPLab Workshop 2013

Prof. Jesús Bernardino Alonso Hernández, Departamento de Seńales y Comunicaciones, Universidad de Las Palmas de Gran Canaria (ULPGC), Las Palmas de Grand Canaria, Spain

Emotional Speech Characterization for Real Time Applications in Real Environments
Traditional methods of emotional discrimination use prosodic and paralinguistic features, which are determined by a linguistic segmentation of the locution. However, these methods are not appropriate in real time applications because of their high computational cost and the linguistic segmentation requirement by locutions. This presentation proposes a new strategy based on a few prosodic and paralinguistic features set obtained from a temporal segmentation of the speech signal. This new strategy is robust to interfering noises that are present in real environments, offering a low computational cost and improving the performance of a segmentation based on linguistic aspects.

Dr. Hicham AtassiDept. of Telecommunications, Brno University of Technology, Czech Republic

Advanced Dialogue Analysis as a Part of an Autonomous Intelligent System for Call Centres Surveillance and Assessment
There is a huge boom of call centres worldwide, with an ever growing number of companies operating through phone marketing. The performance of operators in such subjects is of crucial importance in addition to the clients’ feedback. However, it is very hard to manually evaluate the quality of services provided, or to assess the agents’ performance. For example, if a company has 20 operators working daily for 7 hours, 5 days per week, then the phone calls recorded over one month make about 2800 hours. Since it is impossible to manually check all these phone calls in order to form a reliable image of agents’ performance or to assess the quality of services. In the light of this, it appears evident that some kind of an automatic analysis of speakers’ characteristics is indispensable for phone-marketing and indeed all subjects that involve costumer support services in their structure.

In this presentation, we will present a new autonomous intelligent system for call centres surveillance and assessment. The system is based on advanced speech processing techniques aiming to recognize multiple speakers’ characteristics such as emotional state, age and gender. A special focus will be on the dialogue between the telephone call participants. The results of advanced analysis of spontaneous speech records obtained from real call centres showed interesting correlations between the dialogue basic characteristics and both activation and evaluation levels of speaker’s emotional state. We will also show that the dialogue characteristics can be successfully employed to assess the performance of call centres agents.

Dr. Peter BalazsAcoustics Research Institute, Austrian Academy of Sciences, Austria

Gabor dual windows using convex optimization
The question of finding good synthesis filterbanks is an important topic for signal processing. In the case of redundant Gabor frames an infinite number of dual frames allow perfect reconstruction. We employ convex optimization methods to design dual windows, optimizing various constraints, e.g. good time and frequency decay. Numerical experiments show that alternate dual windows with considerably improved features can be found.

Prof. Ilker BayramElectronics and Communication Engineering Department, Istanbul Technical University, Turkey

Fusion of Audio Signals Affected by Time-Varying Noise
We consider a scenario that involves multiple audio recordings of a source, where each of the recordings is affected by time-varying noise with unknown characteristics. We propose a reconstruction scheme that combines classical beamforming formulations with models based on the sparsity of the source signal’s spectrogram. Specifically, we propose a parametric reconstruction and choose the parameters by solving a convex minimization problem. We demonstrate that the parameters chosen this way are close to the ideal values, if information about the statistics of the noise terms were available. We also discuss how to derive real-time algorithms based on the proposed approach and present a real-life application.

Dr. Radek BenesDept. of Telecommunications, Brno University of Technology, Czech Republic

Image processing for measurement of intima media thickness
The presentation describes an automatic system for the intima media thickness (IMT) measurement performed in B-mode ultrasound images. The presentation will be divided into two main parts according to the division of the system. First part will be focused on the artery localization, where the image processing based on local statistics is used. The segmentation of intima and media layers as well as a a pre-selection of the most suitable places is described in the second part of the presentation.

Dr. Peter DrotarDept. of Telecommunications, Brno University of Technology, Czech Republic

A Novel Biomarker for Evaluation of Parkinson’s Disease: In-Air Movement in Handwriting
Parkinsons disease (PD) is neurodegenerative disorder with very high prevalence rate occurring mainly among elderly. One of the most typical symptoms of PD is deterioration of handwriting that is usually the first manifestation of Parkinsons disease. In this study, a new modality – in-air trajectory during handwriting – is proposed to efficiently diagnose PD. Experimental results showed that analysis of in-air trajectories is capable of assessing subtle motor abnormalities that are connected with PD. Moreover, conjunction of in-air trajectories with conventional on-surface handwriting allows us to build predictive model with PD classification accuracy over 80%. In total, we compute over 600 handwriting features. Then, we select smaller subset of these features using two feature selection algorithms: Mann-Whitney U-test filter and relief algorithm, and map these feature subsets to binary classification response using support vector machines.

Prof. Anna EspositoSecond University of Naples, Caserta and IIASS, Italy

On the Effectiveness of Visual and Vocal Channels in Transmitting Dynamic Emotional Information: a Cross-Cultural Study
The present work aims at examining emotion recognition within and across cultures. It reports on a set of perceptual experiments designed to explore the human ability to identify emotional expressions dynamically presented through visual and auditory channels. Two cross-modal databases of dynamic verbal and non-verbal emotional stimuli based on video-clips extracted from American and Italian movies, respectively, were defined and exploited for the experiments. The databases allows a cross-modal analysis of audio and video recordings with the aims of identifying distinctive, multi-modal and cultural specific emotional features from multi-modal signals, as well as for defining methodologies and mathematical models for the automatic implementation of naturally human-like communication interfaces.

In the first study, American, French, and Italian subjects were involved in a comparative analysis of subjective perceptions of six emotional states dynamically portrayed by visual and vocal cues, exploiting the database of American emotional stimuli. In the second study, American and Italian subjects were tested on their ability to recognize six emotional states through the visual and auditory channel, exploiting the database of Italian emotional stimuli. The aim is to explore whether there exists a difference in the efficacy of the visual and auditory channels to infer emotional information and whether cultural context, in particular the language, may influence this difference. This hypothesis is investigated including as participants in each of the two studies, one group of native speakers of the language, belonging to the same cultural context of the video-clips used as stimuli (i.e., American subjects for the American stimuli in the former study, and Italian subjects for the Italian stimuli in the latter one). Results showed that emotional information is affected by the communication mode and that language plays a role.

Dr. Kamran FarooqDepartment of Mathematics and Computing Science, University of Stirling, Scotland

Efficient clinical risk assessment of cardiovascular patients using an Ontology driven and Machine Learning Approach
The aim of this study is to help improve the diagnostic and performance capabilities of RACPC, by reducing delay and inaccuracies in the cardiovascular risk assessment of patients with chest pain by helping clinicians effectively distinguish acute angina patients from those with other causes of chest pain. Key to our new approach is (1) an intelligent prospective clinical decision support (CDS) framework for primary and secondary care clinicians, (2) learning from missing/impartial clinical data using Bernoulli mixture models and Expectation Maximisation (EM) techniques, (3) encoding of clinical expert’s knowledge using a Bayesian Network representation to handle uncertainties in clinical decision making and (4) utilisation of state-of-the-art pattern recognition and data mining techniques for the development of intelligent risk.

Prof. Marcos Faundez-Zanuy, EUP Mataró, TecnoCampus Mataró-Maresme, Spain

Preliminary experiments on automatic gender recognition based on online capital letters
In this lecture we present some experiments to automatically classify online handwritten text based on capital letters. Although handwritten text is not as discriminative as face or voice, we still found some chance for gender classification based on handwritten text. Accuracies are up to 74%, even in the most challenging case of capital letters.

Prof. Malay Kishore DuttaDept. of Electronics & Communication Engineering, Amity University, Noida, India

Ownership of Digital Signals using Biometric features
With the development of high speed internet and transmission of audio files over the internet, there is a need of copyright protection and digital right management. Illegal reproduction and unauthorized distribution of digital audio has become a high alarming problem in protecting the copyright of digital media. This talk will address a method of embedding a biometric feature based digital signature in the digital media in a perceptually transparent manner. This embedding digital Signature will ensure perceptual transparency is maintained and the attack characterization is done to achieve robustness so that the digital signal survives signal processing attacks. Biometric features like iris image and fingerprint image has been used to generate the digital signature and it has been tested in digital image and digital audio signals. Unique identification has been done for the extracted digital signature which indicates a logical ownership of the digital signal has been established.

Prof. Juan Ignacio Godino-LlorenteDepartamento de Ingeniera de Circuitos y Sistemas, Universidad Politcnica de Madrid, Spain

Bioengineering and Optoelectronics Group of the Universidad Politécnica de Madrid
The aim of this presentation is to introduce the research of Bioengineering and Optoelectronics Group at the Universidad Politécnica de Madrid. The main fields of interest, the structure of group and possible topics of cooperation will be presented.

Robust automatic tracking of the vocal folds
The present work describes a new methodology for the automatic tracking of the glottal space from high-speed digital images of the larynx. This approach involves four steps: Firstly, the region with maximal glottal opening is found, by detecting the total intensity variation of the video in the x and y axis. Secondly, we create a standard template using the information obtained from manual segmentations. Thirdly, using normalized cross correlation the surface that represents the best matching between the template and the next frame is obtained. The matching area will be the initialization for the segmentation algorithm in each frame. Finally, using an algorithm based on active contours we obtain the glottal gap. This procedure is done iteratively until the last frame has been reached. The performance, effectiveness and validation of the approach is demonstrated even in high-speed recordings in which the images present an inappropriate closure of the vocal folds.

Dr. Nicki HolighausAcoustics Research Institute, Austrian Academy of Sciences, Austria

Pave the way with circles: Efficient algorithms for the sampled short-time Fourier transform on nonseparable lattices
Lattices are subgroups of the time-frequency plane. We introduce the concept of sampling the short-time Fourier transform (STFT) on a lattice and its potential benefits compared to the classical rectangular (aZ x bZ) sampling scheme. Once the basic concepts are in place, we discuss several efficient methods to reduce the computation of STFT analysis and reconstruction on arbitrary lattices to well-known efficient algorithms for rectangularly sampled transforms.

Dr. Radovan JiříkInstitute of Scientific Instruments of the ASCR, Academy of Sciences of the Czech Republic, Czech Republic

Data acquisition and analysis in MRI and ultrasound perfusion imaging

Dr. Jan KarásekDept. of Telecommunications, Brno University of Technology, Czech Republic

Optimization of processes in warehouse environments based on evolutionary computations
The presentation introduces a combination of two scheduling problems. The first problem is related to the well-known Job Shop Scheduling Problem applied in logistic warehouses and distribution centers. The second problem is related to Automated Guided Vehicle Systems and the problem of vehicle routing. The combination of these two problems is advantageous in planning and scheduling the jobs in warehouses and prevents blocking and vehicle collisions. The main contributions of the presented work are a) to define the problem of half-automated logistic companies, b) to define a set of benchmarks, c) to define a simple evaluation criterion, and d) to give a baseline results for further research using genetic programming.

Prof. Walter G KropatschPattern Recognition and Image Processing Group, Institute of Computer Graphics and Algorithms, Vienna University of Technology, Austria

A Hierarchical Representation for Vision
Vision involves huge amounts of measurements which need to be processed in extremely short time. Pyramidal structures and fast shrinking stacks of plane graphs have the potential to cope with this challenge. We will review the basic structures and address the major recent result, the preservation of topology and our current research, their use for top-down attention and selective spatio-temporal sampling.

Dr. Ales KrupkaDept. of Telecommunications, Brno University of Technology, Czech Republic

Segmentation of Sedimentary Grain in Electron Microscopy Image
The presentation is dedicated to the research concerning automatic processing of sedimentary grains captured by electron microscope. Specifically, it describes the developed segmentation method. The method utilizes the approach of region splitting and merging. In the splitting stage, the marker-based watershed segmentation is used. In the merging phase, the typical characteristics of grains in electron microscopy images are exploited for proposing special metrics, which are then used during the merging stage to obtain a correct grain segmentation. The results of experiments as well as the direction of future research will be discussed.

Prof. Miren Karmele López de Ipina Pena, Department of System Engineering and Automation, University of the Basque Country, San Sebastián, Spain

On the selection of biomarkers from non-invasive writing signals, applied to Essential Tremor
This project, is part of a cross wide-ranging study for the diagnosis of Essential Tremor (Donostia University Hospital/ The Institute for Health Research Biodonostia). The main objective is to characterize tremor, from clinical and genetic standpoint in each of the families with familial essential tremor and identify common features for each of the mutations / genetic loci identified.

Specifically in this project, we are developing a system for automatic selection of biomarkers from drawings and handwriting. Moreover, it will be also analyzed the presence of integrated features of other diseases such as stress. Then these new biomarkers will be integrated with those obtained in the Biodonostia’s study

It should be highlighted that the use of this technology could provide non-invasive undoubted benefits towards the development of more sustainable, low cost and high quality technologies. These systems are easily adaptable to user and environment and from a social and economic point of view very useful in real complex environments.

Jiří MekyskaDept. of Telecommunications, Brno University of Technology, Czech Republic

New Approaches in the Pathological Speech Signal Analysis
Speech signal processing is one of the popular non-invasive technique of the vocal tract or vocal cords pathology analysis. A possibility to objectively quantify different speech dysfunctions is a step for a better and more efficient diagnosis. The aim of this lecture is to describe the whole process of the pathological speech signal analysis with a special focus on feature extraction. There will be described the most common speech dysfunctions, suitable speech tasks, local speech features (basic features, non-linear dynamic features, features based on empirical mode decomposition EMD), global speech features (description of tongue movement), high-level speech features (some statistics), different approaches of feature selection and possible applications of these techniques in a medical and pharmaceutical industry.

Dr. Sebastian Land, Rapid-I, Germany

„Quantify your life“ – How quantitative analysis changes science
In the past years both computational power and data storage capacity became available in abundance. Nowadays everybody carries a device in his pocket that can not only fly a rocket to the moon, but could control an armada of rockets, decode an incredibly high resolution movie to entertain the crew and still would idle to a good extend. This has enabled the emergence of data driven sciences, that use this computational power to search for patterns in tons of data and turn them into usable knowledge. We will demonstrate how data driven science enables everyone to be a Genius of the size of Archimedes and will give examples where these techniques are applied to ease our lives.

Dr. Clarice PoonCentre for Mathematical Sciences, University of Cambridge, United Kingdom

Beyond incoherence and beyond sparsity: compressed sensing in the real world
The current theory of compressed sensing asserts that if we have sparsity with respect to some dictionary, which is incoherent with our measurement system, then the original signal may be recovered by collecting samples uniformly at random in accordance to the sparsity. However, in a lot of important applications, such MRI and reflection seismology, this assumption of incoherence is false and the existing theory does not apply. Yet, empirical studies have shown that compressed sensing can be successful applied via variable density sampling strategies. How can we theoretically justify this?

In this talk, I will generalize the two key ingredient of incoherence and sparsity to asymptotic incoherence and asymptotic sparsity respectively, and demonstrate that these new notions do exist in practice. Under these more realistic assumptions, I will introduce a new theoretical framework which can explain the success of compressed sensing in practical applications. I will also discuss how this framework can be applied for a variety of sparsifying dictionaries, including wavelets, shearlets and total variation. A direct implication of this theory is that firstly, the success of compressed sensing is resolution dependent, and secondly, the optimal sampling strategy is signal dependent. Thus, from a practical viewpoint, a clear understanding of these ideas is important to successful implementation of compressed sensing.

Dr. Zdeněk PrůšaAcoustics Research Institute, Austrian Academy of Sciences, Austria

Filterbanks and block-processing in LTFAT
In the talk, I will give an overview of the current state of the Large Time-Frequency Analysis Toolbox (LTFAT), which is modern Matlab/Octave toolbox for working with time-frequency analysis and synthesis. I will focus on the new common filterbank backend which serves as a base for many types of transforms and on the block-processing framework capable of real-time processing directly in Matlab/Octave.

Dr. Benjamin RicaudSignal Processing Laboratory 2, École polytechnique fédérale de Lausanne EPFL, Switzerland

Gabor and Wavelet transforms for signals defined on graphs
Wavelet and Gabor transformations are among the most popular tools for the analysis of audio signals or images. Nowadays, more and more data problem are modeled by signal on graphs (a signal on the graph is a function associating a value to each node of the graph). Hence a graph version of the wavelet and Gabor transformation would be of great interest for these kinds of graph structured problems. However, their generalization to the graph setting is not straightforward. In order to construct these transformations we need a notion of translation, modulation and dilation of a window defined on the graph. We will propose a definition of these notions which fit the particularities of the graph settings and see how we can construct such transformations.

Dr. Maria Teresa RivielloSecond University of Naples, Caserta and IIASS, Italy

Language and Gender Effect in Decoding Emotional Information
Identifying emotions in human interactions is a complex task: a) the encoding/decoding processing of emotional information occurs through several communication modalities simultaneously (facial, vocal and gestural expressions, among others); b) numerous factors influence the way emotions are expressed and interpreted, comprising cultural and individual characteristics.

This work attempts to account for the impact of cultural, language and gender factors in affecting the decoding process of emotional information, also considering the role of the communication modality (audio, mute video and audio-video) through which it is portrayed. In particular, it explores the ability of Lithuanian male and female participants in decoding emotional information exploiting male and female expressions. The emotional data used in the experiments consist of realistic, dynamic and mutually related audio, mute video and audio-video stimuli. The stimuli were extracted from American English, as a globally spread language, and Italian, as a country specific language, movie scenes. The main goal is to investigate whether the decoding processing of the emotional states in the three considered expressive contexts (audio, mute video and audio-video) depends on the familiarity with the language and the gender of both participants (receivers) and actors (senders). Results show that Lithuanian subjects recognition accuracy is affected by the language specificity of the stimuli. Familiarity with the language also seems to play a role in the assessment of gender differences in decoding emotions. Significant gender effects, in fact, were exclusively found when Lithuanian subjects examined Italian vocal expressions.

Dr. František RundDept. of Radio engineering, Czech Technical University in Prague, Czech Republic

Virtual Acoustic Space as Assistive Technology
Virtual Acoustic Space can be rendered using headphones and signal processing. Apart from application in computer games, multimedia etc., it is considered for application in field of Assistive Technologies, e.g. for orientation of visually impaired people. The aim of the contribution is to show our approach and review the associated problems.

Prof. Jordi Solé i Casals, Department of Digital Technologies and Information, Polytechnic School, University of Vic, Spain

Digital Technologies Research Group at the University of Vic
The aim of this presentation is to introduce the research of Digital Technologies Research Group at the University of Vic. The main fields of interest, the structure of group and possible topics of cooperation will be presented.

Diagnosis of Alzheimer’s disease from EEG using machine learning algorithms
Alzheimer’s disease (AD) is a neurodegenerative disease and the most common form of dementia. It is a progressive and irreversible deterioration of brain functions that starts with loss of memory and leads to other cognitive impairments, such as deficits of language or deficits of judgment. Recently, significant advances have been made in the early diagnosis of AD from EEG, yet it remains a difficult task. In the present study we investigate the early diagnosis of AD using synchrony measures and frequency Relative Power on EEG signals. First we explore the use of a single feature in order to compute classification rate, looking for the best frequency range. Then, a multiple feature classification system is proposed that improves all previous results: 95% of classification rate for MCI data set using 11 features; 100% of classification rate for Mild AD data set using four features.

Dr. Alois SontacchiInstitute of Electronic Music and Acoustics , Universität für Musik und darstellende Kunst in Graz, Austria

Principles and applications of 3D reproduction using Ambisonics
Sound field reproduction has become important for simulation tasks in scientific disciplines. Sound field description using spherical harmonics leads to the so called Ambisonics approach. Recently improvements in regard to applicability and understanding have made its strength more apparent. This rendering approach can be applied both for headphones and loudspeakers. It provides advantages such as system scalability and modularity, as well as treatment of irregular loudspeaker layouts on an in whole or in part surrounding sphere.

An introduction in sound field reproduction using Ambisonics will be given. Examining the underlying principles the focus is put on application scenarios. In general, circular and spherical loudspeaker arrangements are addressed. However, the contemporary theory of spherical harmonics does not require such restrictions and offers access to arbitrary layouts to a certain extend.

Applications for headphones and loudspeakers are treated. Objective quality criteria are revisited and proposed psycho-acoustic motivated improvements are discussed, in order to optimize the perceived audio reproduction.

Sound field reproduction using Ambisonics provides benefits in implementations. Reconstruction quality can be analytically predicted in regard to physical quantities as well as acoustically relevant perceived measures. Therefore, overall expense and sound field reproduction quality can be scaled in accordance to the considered application.

Dr. Michal ŠorelDept. of Image Processing, Institute of Information Theory and Automation (UTIA), Academy of Sciences, Czech Republic

Fast algorithms for Bayesian JPEG decompression
JPEG decompression can be formulated as a probabilistic problem and solved in the standard way using the Bayesian approach. The choice of image prior probability distribution influences complexity of the corresponding minimization problem. In this talk we show how a convenient form of this prior allows for efficient solution by a primal-dual method.

Dr. David SvobodaCentre for Biomedical Image Analysis, Masaryk University, Czech Republic

On generating benchmark datasets for evaluation of segmentation and tracking algorithms in fluorescence microscopy
In fluorescence microscopy, the proper evaluation of image segmentation and tracking algorithms is still an open problem. As the ground truth for cell image data (and measurements on them) is not available in most experiments, the outputs of different image analysis methods can hardly be verified or compared to each other.

We created a toolbox that can generate 3D digital phantoms of specific cellular components along with their corresponding images degraded by specific optics and electronics. The images can represent static scenes (fixed cell) as well as time-lapse sequences (living cells). Such synthetically generated images can serve as a benchmark dataset for measuring the quality of various segmentation and tracking algorithms.

Dr. Sakari TervoDepartment of Media Technology, Aalto University School of Science, Finland

Parametric approaches to analysis of spatial room impulse responses
Room acoustics is often evaluated from omni-directional measurements and parameters calculated from them do not adequately describe the details of acoustics. Especially the spatial features of the acoustics are lost in the process. Recently, array signal processing has been applied to the analysis of room acoustics. This allows the inspection of the acoustics in spatiotemporal as well as in time-frequency domain. This presentation discusses the parametric array processing algorithms that we have applied for the analysis of concert hall acoustics. In detail, the limitations of the algorithms with respect to the physical model of the spatial room impulse response are reviewed. In addition, visualizations of the results for several concert halls are shown.

Prof. Carlos Manuel Travieso González, Departamento de Seńales y Comunicaciones, Universidad de Las Palmas de Gran Canaria (ULPGC), Las Palmas de Grand Canaria, Spain

Emotional Load Detection based on the mouth opening
In this work has been developed a tool that extracts and evaluates the emotional load of a person from a facial image through movement of the lips. This opening of the mouth is used to measure and characterize it. This parameter represents an important tool for Neurologists to detect emotion from a patient and follow its evolution for any emotional or psychological disorder, and compared to the evolution of treatment. This system is able to establish four states: silence, low emotion, normal emotion, high emotion, but at the same way, it shows the quantification for each detected state.

Dr. Michal TrzosDept. of Telecommunications, Brno University of Technology, Czech Republic

Representation of frequency modulated audio signals using Fast harmonic transform
Spectral analysis of audio signals using the quasi-harmonic model without frequency modulation is usually carried out using Short-time Fourier transform with sufficient precision. However if frequency modulation is present in the analysed signal, spectral lines of the STFT are no longer aligned with the fundamental frequency change which causes estimation error of the harmonic component of the sound. In Harmonic transform, the transformation kernel is aligned with the fundamental frequency change which allows more accurate analysis. This presentation introduces fast version of the Harmonic transform with some possible applications.

Dr. Jan VybiralDepartment of Mathematics, Technical University Berlin, Germany

Survey on Compressed Sensing and Applications
Compressed Sensing is a new technique of signal processing, which allows to exploit sparsity or compressibility of a signal with respect to some known basis or dictionary and which allows to design extremely effective non-adaptive sampling algorithms. This breakthrough led to a shift in a traditional view of signal acquisition in electrical engineering. This survey talk will provide an overview of the most important mathematical ideas behind this new sampling theory, including the notions of null space property, restricted isometry property, and sparse recovery by l1-minimization. We will also discuss connections of this field with other disciplines like stochastic, numerics, linear algebra, and functional analysis. Furthermore, we will present how these ideas can be further developed and applied to Matrix Completion problem or Phase Retrieval problem. We will also discuss the applications of Compressed Sensing to the design of new generation of sensing devices and map on the progress in this area.

Dr. Christoph WiesmeyrNuHAG, University of Vienna, Austria

*-lets for everybody: Define your own transform
For the analysis of one-dimensional signals the literature provides many different transforms. Among the most popular and well known are the Gabor transform and wavelets. Recently, the ERBlets have been introduced, which provide a transform adapted to human hearing. While Gabor atoms are are designed for a linear frequency scale, wavelets follow a logarithmic and ERBlets the ERB frequency scale. We will discuss a unified framework to realize transforms adapted to all the mentioned frequency measures with perfect reconstruction. The approach can be used to realize any given frequency scaling leading to a rich world of transforms.