The Audio Analysis Lab is organizing a Ph.D. course on Distributed Signal Processing at Aalborg University in the Spring of 2014. The course will be given by Richard Heusdens who is Guest Professor at Aalborg University.

Description: In the course Distributed Signal Processing, attention will be paid to decentralized signal processing techniques. Due to the explosion in size and complexity of modern datasets, it is increasingly important to be able to solve problems with a very large number of features or training examples. In industry, this trend has been referred to as ‘Big Data’, and it has had a significant impact in areas as varied as artificial intelligence, internet applications, computational biology, medicine, finance, marketing, journalism, network analysis, weather forecast, telecommunication, and logistics. As a result, both the decentralized collection or storage of these datasets as well as accompanying distributed solution methods are either necessary or at least highly desirable. In this course we will focus on two signal processing techniques for decentralized processing: one based on graphical models and one based on convex optimization. We will consider the following topics: graphical models, probabilistic inference, message passing, min-sum/max-product algorithm, Jacobi and Gauss-Seidel algorithm, convex optimization, gossip algorithm, dual ascent, dual decomposition and alternating direction method of multipliers (ADMM).

The course will run April 4-8. If you are interested in following the course, sign up at http://phd.moodle.aau.dk/ or https://phdcourses.dk/ or send course organizer Mads Græsbøll Christensen an email at mgc@create.aau.dk.

Our new book, entitled Signal Enhancement with Variable Span Linear Filters, is now out. It is published by Springer and is authored by Jacob Benesty, Jesper Rindom Jensen, and Mads Græsbøll Christensen. The book continues our work on a unified subspace and optimal filtering view on noise reduction.

Abstract:

This book introduces readers to the novel concept of variable span speech enhancement filters, and demonstrates how it can be used for effective noise reduction in various ways. Further, the book provides the accompanying Matlab code, allowing readers to easily implement the main ideas discussed. Variable span filters combine the ideas of optimal linear filters with those of subspace methods, as they involve the joint diagonalization of the correlation matrices of the desired signal and the noise. The book shows how some well-known filter designs, e.g. the minimum distortion, maximum signal-to-noise ratio, Wiener, and tradeoff filters (including their new generalizations) can be obtained using the variable span filter framework. It then illustrates how the variable span filters can be applied in various contexts, namely in single-channel STFT-based enhancement, in multichannel enhancement in both the time and STFT domains, and, lastly, in time-domain binaural enhancement. In these contexts, the properties of these filters are analyzed in terms of their noise reduction capabilities and desired signal distortion, and the analyses are validated and further explored in simulations.

Read more about the book here  http://www.springer.com/us/book/9789812877383

A number of papers (co-)authored by lab members will be presented at EUSIPCO 2015 in Nice, France:

• Multi-Pitch Estimation and Tracking Using Bayesian Inference in Block Sparsity
• Pitch Estimation of Stereophonic Mixtures of Delay and Amplitude Panned Signals
• A Fast Algorithm for Maximum Likelihood-based Fundamental Frequency Estimation
• A Discriminative Approach for speaker Selection in Speaker De-identification Systems

Two papers will also be presented at InterSpeech 2015 in Dresden, Germany:

• Enhancement of Non-Stationary Speech using Harmonic Chirp Filters
• Least Squares Estimate of the Initial Phases in STFT based Speech Enhancement

We are very pleased to announce that the Audio Analysis Lab’s grant proposal entitled Signal Processing for Diagnosis of Parkinson’s Disease from Noisy Speech has been accepted by the Danish Council for Independent Research | Technology and Production Science! The lab will receive 6.5 million DKK for the project, which will employ two Ph.D students and one postdoc.

Many people are affected by Parkinson’s disease (PD) in some way. There currently exists no cure, there are no known biomarkers that can be used for diagnosis, and the number of people with PD is expected to rise dramatically in the near future. The project aims at finding accurate and robust signal processing methods for analyzing natural, noisy speech for early diagnosis and monitoring of the progression of PD based on parametric models. The project is a international collaboration between Aalborg University, MIT, University of Colorado-Boulder, the Parkinson’s Voice Initiative, and Lund University.

You can read more about the granted projects here.

We are happy to announce that Richard Heusdens from Delft University of Technology is now Guest Professor at Aalborg University and will be visiting us a number of times. He will work with us on distributed signal processing for ad hoc camera and microphone networks and will be giving a Ph.D. course on distributed signal processing. The professorship is kindly sponsored by the VELUX Visiting Professor Programme 2015-2016.

Richard Heusdens is an Associate Professor at the Department of Mediamatics of the Delft University of Technology. He received his HTS (institute of technology) diploma from the HTS Alkmaar in 1988. He received the M.Sc. and Ph.D. degrees from Delft University of Technology, Delft, The Netherlands, in 1992 and 1997, respectively. Since 2002, he has been an Associate Professor in the Department of Mediamatics, Delft University of Technology. In the spring of 1992, he joined the digital signal processing group at the Philips Research Laboratories, Eindhoven, The Netherlands. He has worked on various topics in the field of signal processing, such as image/video compression and VLSI architectures for image processing algorithms. In 1997, he joined the Circuits and Systems Group of Delft University of Technology, where he was a Postdoctoral Researcher. In 2000, he moved to the Information and Communication Theory (ICT) Group, where he became an Assistant Professor responsible for the audio and speech processing activities within the ICT group. He is involved in research projects that cover subjects such as audio and speech coding, speech enhancement, and digital watermarking of audio. Research projects he is involved in cover subjects such as audio and speech coding (Sicas, ARDOR, ASC), speech enhancement (SpEnt), and digital watermarking of audio (DIWAMETRICS)

Audio Analysis Lab members are currently in Brisbane, Australia attending ICASSP 2015. We are well-represented at this years’ conference with the following papers authored by lab members:

• ON FREQUENCY DOMAIN MODELS FOR TDOA ESITMATION
• PITCH ESTIMATION AND TRACKING WITH HARMONIC EMPHASIS ON THE ACOUSTIC SPECTRUM
• PITCH AND TDOA-BASED LOCALIZATION OF ACOUSTIC SOURCES WITH DISTRIBUTED ARRAYS
• PSEUDO-COHERENCE-BASED MVDR BEAMFORMER FOR SPEECH ENHANCEMENT WITH AD HOC MICROPHONE ARRAYS
• A JOINT AUDIO-VISUAL APPROACH TO AUDIO LOCALIZATION

Moreover, during the opening and award ceremony, AAU alumni Daniele Giacobello was celebrated as he received an IEEE Signal Processing Society Young Author Best Paper Award for the following journal paper:

Daniele Giacobello, Mads Græsbøll Christensen, Manohar N. Murthi, Søren Holdt Jensen and Marc Moonen, “Sparse Linear Prediction and Its Applications to Speech Processing“, IEEE Transactions on Audio, Speech, and Language Processing, Volume: 20, No. 5, July 2012.

On April 7, our annual workshop, which is now called Audio Analysis Workshop, was held in Aalborg. It is sponsored by the Villum Foundation via our YIP grant. The program consisted of 17 talks on topics such as pitch estimation, speech enhancement, microphone arrays, audio applications of sparse approximations, and fast implementations, and 15 people participated in the workshop. We thank everybody for their contributions.

We are extremely proud to announce that AAU alumni Daniele Giacobello has received an IEEE Signal Processing Society Young Author Best Paper Award for the following journal paper, which was part of his Ph.D. work while at AAU:

Daniele Giacobello, Mads Græsbøll Christensen, Manohar N. Murthi, Søren Holdt Jensen and Marc Moonen, “Sparse Linear Prediction and Its Applications to Speech Processing“, IEEE Transactions on Audio, Speech, and Language Processing, Volume: 20, No. 5, July 2012.

The paper was co-authored by lab member Mads Græsbøll Christensen who also supervised Daniele at AAU with Søren Holdt Jensen. The work has received a high number of citations in just two years (55 according to Google Scholar, and more than 150 when counting the preceding conference papers), has been accessed more than 1000 times in IEEE Xplore, and has spawned research in several different fields, including audio processing, seismography, video processing, and general analysis of time series and filter design.

The Young Author Best Paper Award honors the author(s) of an especially meritorious paper dealing with a subject related to the Society’s technical scope and appearing in one of the Society’s solely owned transactions or the Journal of Selected Topics in Signal Processing and who, upon the date of submission of the paper, is less than 30 years of age.

Read more about the IEEE SPS 2014 Awards at http://www.signalprocessingsociety.org/awards-fellows/award-recipient/.

Read more about the IEEE SPS Young Author Best Paper Award at http://www.signalprocessingsociety.org/awards-fellows/awardspage/youngauthor/.

As part of his trip to California to attend the Asilomar conference, Mads Græsbøll Christensen visited Beats by Dr. Dre (owned by Apple) in Los Angeles, CA, on November 10 and gave a talk there on the Audio Analysis Lab’s work on parametric and statistical methods fro speech processing. Beats by Dr. Dre makes premium consumer headphones, earphones, and speakers. AAU alumni and former student of Audio Analysis Lab head Mads Græsbøll Christensen, Daniele Giacobello, works at Beats by Dr. Dre.

On November 6 2014, Mads Græsbøll Christensen of the Audio Analysis Lab, as part of his trip to the Asilomar conference, visited Google in Mountain View, CA and gave a talk there about the lab’s work on parametric and statistical methods for speech processing. Google is, of course, well-known for their search engine technology, Android and ChromeOS, technologies that also involve a lot of audio processing to facilitate voice communication, language processing, and machine listening.