Author: Mads Græsbøll Christensen

On Sunday morning August 20, Audio Analysis Lab members Jesper Kjær Nielsen, Jesper Rindom Jensen, and Mads Græsbøll Christensen gave a three-hour tutorial at INTERSPEECH in Stockholm, Sweden. The tutorial was entitled Statistical Parametric Speech Processing: Solving Problems with the Model-based Approach and included a lot of the lab’s work in recent years on model-based methods for processing of speech and audio signals, including also our work on speech modeling for diagnosis of Parkinsons. The tutorial was popular with 70+ participants signed up! More info about the tutorial can be obtained here. INTERSPEECH 2017 emphasizes an interdisciplinary approach covering all aspects of speech science and technology spanning basic theories to applications.

On August 24 2017, the Audio Analysis Lab’s annual workshop, called the Audio Analysis Workshop, was held at AD:MT in Rendsburggade 14 in Aalborg. This year’s edition is sponsored by the Independt Reserach Fund Denmark via our project on diagnosis of Parkinsons disease from voice signals. There were 25 participants from the lab, industrial partners, and universities abroad. The workshop featured 18 talks on ongoing research on topics such as frequency estimation, music signal modeling, distributed signal processing, Parkinsons, noise reduction, hearing aids, array processing, and speech intelligibility prediction. This was the sixth edition of the workshop which started in 2012.

The first results of the project will be presented next week at ICASSP 2017. The 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2017 is being held March 5-9, 2017 in New Orleans, USA. It is the top signal processing conference in the world. The paper to be presented is written by Liming Shi who is Ph.D. student on the project and is co-authored by Jesper Rindom Jensen and Mads Græsbøll Christensen.

Paper Code: AASP-P11.7	Paper Number: 1626
Title: LEAST 1-NORM POLE-ZERO MODELING WITH SPARSE DECONVOLUTION FOR SPEECH ANALYSIS
Track: AASP: Audio and Acoustic Signal Processing
Session: Audio Quality, Coding and Processing
Time: Thursday, March 9, 13:30 – 15:30
Authors: Liming Shi, Jesper Rindom Jensen, Mads Græsbøll Christensen

We are happy to report that the project is now running at full steam. All the researchers (Ph.D. students, postdocs) employed by or affiliated with the project are now working on their respective topics. These are the following:

• Mahmoud Fakhry (Postdoc, Aalborg University) works on statistical methods for enhancing speech signals for improving the feature extraction for detection of Parkinson’s disease.
• Liming Shi (Ph.D. Student, Aalborg University) works on modeling of speech production and robust estimation from noisy recordings with the aim of extracting physically meaningful features.
• Amir Poorjam (Ph.D. Student, Aalborg University) works on automatic assessment of the quality and pre-processing of speech signals in audio archives.
• Alfredo Esquivel (Ph.D. Student, Aalborg University) works on a speech analysis toolbox for studying speech pathology, et.c., based on parametric methods and speech decompositions.
• Filip Elvander (Ph.D. Student, Lund University) works on methods for reconstruction of speech signals that have been damaged due to nonlinear phenomena such as packetlosses, compression, and clipping, using sparse methods.

We wil publish the results of our research here on the homepage and at the leading conferences (ICASSP, EUSIPCO, InterSpeech) and in the top journals (IEEE/ACM Trans. Audio, Speech, and Language Processing) along the way.

On August 19 2016 the annual Audio Analysis Workshop was held. This year’s edition was co-sponsored by the Danish Council for Independent Research and served also as a kick-off meeting for the project. It featured 14 scientific talks and 2 keynote talks with 18 participants from Lund University, Aalborg Unviversity, Delft University of Technology, GN Resound, and Ashton University. The workshop had a Parkinson’s disease flavored theme with two keynote talks on the topic Parkinson’s disease, how it affects the voice, and how it can be detected from the voice. The first keynote talk, entitled Braak´s hypothesis and its impact on research and treatment in Parkinson´s disease was given by neurologist Lorenz Oppel, Aalborg University Hospital. The second keynote talk was given by Dr. Max Little, Ashton University, and was entitled Algorithms for feature extraction in voicebased analysis of Parkinson’s disease. In his talk, Max gave an overview of his many years of research on the topic. The remaining scientific talks were on different topics, including fast implementations, microphone arrays, music analysis, speech intelligibility, multi-pitch estimation, sparse approximations, classification, and speech enhancement.

Our project has received a lot of interest, and people frequently ask us about it. There is a very nice TED Talk by Prof. Max Little about the ideas behind our project. Max Little has worked on the machine learning for diagnosis from voice signals for many years now, and he is also part of our project. In the video below, he explains the ideas and how it works.

We are currently looking for candidates for the open position as postdoctoral researcher for this project. We are looking for recent graduates with a Ph.D. in signal processing, speech processing, machine learning, or something like that. If anybody’s interested in the position, please contact Prof. Mads Græsbøll Christensen by sending an email to mgc@create.aau.dk. We will post an official call for applications once we know we have qualified candidates.

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 an international collaboration between Aalborg University, MIT, University of Colorado-Boulder, the Parkinson’s Voice Initiative, and Lund University. The project is funded by the Danish Council for Independent Research | Technology and Production Science which has granted 6.5 million DKK for the project.