Welcome to Signal Processing for Hearing Assistive Devices

**Description: **Hearing assistive devices are ubiquitous. They include, for example, devices such as headsets for speech communication in noisy environments (air plane crews, emergency/rescue teams, combat soldiers, police forces, etc.), and hearing care systems, e.g. hearing aids and cochlear implants.

The course consists of lectures and exercises, to emphasize the key points of the lectures. The course has four parts. The first part lays the foundation for the rest of the course, covering fundamental topics such as auditory perception (normal and impaired hearing) and a discussion of the basic principles of hearing assistive devices. The second part provides an overview of basic signal processing problems encountered in hearing assistive devices, and an in-depth treatment of state-of-the-art solutions. These include methods for beamforming and noise reduction, dereverberation, feedback cancellation, hearing loss compensation, etc. Furthermore, an overview is given of methodologies for evaluating hearing assistive devices with a particular focus on methods for intelligibility assessment and estimation. The third part of the course focuses on cochlear implants and diagnostic methods, e.g., methods for screening of hearing in newborns. Finally, the fourth part of the course presents emerging technologies for hearing assistive devices, including machine learning techniques for speech denoising, and signal processing techniques using the emerging wireless infrastructure.

**Prerequisites: **Basic knowledge of statistical signal processing, stochastic processes, and linear algebra. Familiarity/handy with Matlab, Python, or similar.

**Course Fee: **1500 DKK for PhD students and 8000 DKK for industrial participants (lunches, coffees, one course dinner)

**Organizer: **Professor Jesper Jensen, Aalborg University, email: jje@es.aau.dk

**Lecturers: **Associate Prof. Jan Østergaard, Aalborg University, Associate Prof. Zheng-Hua Tan, Aalborg University, Prof. Jesper Jensen, Aalborg University, Dr. James Harte, Interacoustics (invited talk), Dr. Søren Riis, Oticon Medical (invited talk), Prof. Thomas Lunner, Linköping University (invited talk), Prof. Mike Brookes, Imperial College London (invited talk), and Associate Professor Emanuël Habets, Int. Audio Labs Erlangen (invited talk)

**ECTS:** 3.5

**Time:** 6 - 10 November 2017

**Place:**

**Zip code: **

**City: **Aalborg

**Number of seats: **50

**Deadline: **6 September 2017

**Important information concerning PhD courses** We have over some time experienced problems with no-show for both project and general courses. It has now reached a point where we are forced to take action. Therefore, the Doctoral School has decided to introduce a no-show fee of DKK 5,000 for each course where the student does not show up. Cancellations are accepted no later than 2 weeks before start of the course. Registered illness is of course an acceptable reason for not showing up on those days. Furthermore, all courses open for registration approximately three months before start. This can hopefully also provide new students a chance to register for courses during the year. We look forward to your registrations.

- Teacher: Asger Heidemann Andersen
- Teacher: Jesper Jensen
- Teacher: Jan Østergaard
- Teacher: Zheng-Hua Tan

Welcome to The Estimation of Frequency

**Description: **Periodic phenomena occur naturally. Often the periods are obvious. For example, hourly temperature data are expected to have approximate periods of length 24 and 24*365.25 hours approximately. These periods are only approximate, as there is always going to be some variation that is impossible to model deterministically. Often the periodicity or frequency is unknown, and its value crucial to the understanding of the phenomenon. It therefore needs to be estimated from data.

The estimation of frequency from data had occupied scientists since the eighteenth century. The periodogram was introduced by Schuster in 1898 as a means of estimating a `hidden' periodicity, but other techniques were available by then. In particular, the Buys-Ballot method was introduced in 1847. Indeed, Gauss developed an FFT-like technique in about 1805. The statistical theory of the properties of estimators appears to have been evaluated first by Whittle in 1952, who noted that the relevant Cramer-Rao lower bound for frequency was of order T^{-3}, rather than the usual T^{-1}, where T is the sample size. Walker (1971) and Hannan (1975) followed up with rigorous results and proofs for the case where the additive noise has very general properties.

The publication of the Cooley and Tukey FFT algorithm in 1965 has undoubtedly been responsible for the enormous interest in `frequency domain' techniques - those based on the Fourier transforms of data rather than the data themselves. Large numbers of articles have appeared, mainly in the signal processing literature, but also in statistical journals.

In this course, we shall examine many different frequency estimation techniques and establish the asymptotic properties of the estimators. All of the probabilistic and estimation theory needed will be introduced to establish the (strong) consistency and central limit theorems of the estimators. We shall also consider the computational aspects of the algorithms, and use Matlab code to implement them.

**Prerequisites**: A basic knowledge of mathematics as obtained through undergraduate engineering studies.

**Organizer: **Professor Mads Græsbøll Christensen, email: mgc@create.aau.dk

**Lecturer: **Professor Barry Quinn

**ECTS:** 3

**Time: **4. - 8. September 2017

**Place: **Rendsburggade 14, 9000 Aalborg

4 September room 4.531

5 September room 5.125

6, 7 and 8 September room 5.441

**Zip code: **

**City: **Aalborg

**Number of seats: **25

**Deadline:** 15. August 2017

**Important information concerning PhD courses** We have over some time experienced problems with no-show for both project and general courses. It has now reached a point where we are forced to take action. Therefore, the Doctoral School has decided to introduce a no-show fee of DKK 5,000 for each course where the student does not show up. Cancellations are accepted no later than 2 weeks before start of the course. Registered illness is of course an acceptable reason for not showing up on those days. Furthermore, all courses open for registration approximately three months before start. This can hopefully also provide new students a chance to register for courses during the year. We look forward to your registrations.

- Teacher: Mads Græsbøll Christensen
- Teacher: Barry Quinn
- Teacher: Barry Gerard Quinn

Welcome to Control and Optimization

**Description: **Optimal control is the problem of finding control for a dynamic system such that a certain performance function is minimized. The subject stems from the calculus of variations. The prompt development of optimal control in 1950s owns two inventions: the maximum principle by L.S. Pontryagin and dynamic programming by R. Bellman. Today, the stress is on developing efficient numerical methods for solving a class of optimal control problems (herein convex optimization). Optimal control finds its application not only in engineering, but also in economics, biology and logistics.

The course has two main parts, which in headlines are: Optimal Control, Optimization and Applications.

In the course, we will concentrate on the foundation of optimal control. We will discuss necessary and sufficient condition for optimality, and various types of constraints. We will address the question of existence of optimal strategies.

We cover two main results in optimal control theory, the Bellman (or Hamilton-Jacobi-Bellman) equation and the (Pontryagin) maximum principle. We show how the dynamic programming principle works for an optimal control problem by using the Bellman equation to solve linear quadratic control problems. Moreover, we apply the maximum principle to linear quadratic control problems. The second part of the course will be devoted to the optimization algorithms and the application of optimal control. Here, we will concentrate on convex optimization techniques: conic optimization, dual decomposition, admm, as they provide tangible methods for solving (convex) optimization problems. We will discuss software for optimization of dynamical systems.

**Prerequisites**: A basic knowledge of mathematics as obtained through undergraduate engineering studies.

**Organizer: **Associate Professor John Leth.

**Lecturers: **Dr. Joachim Dahl, MOSEK ApS, Associate Professor Jan Østergaard, Department of Electronic Systems, Associate Professor John Leth, Department of Electronic Systems, Associate Professor Christoffer Sloth, Department of Electronic Systems, and Professor Rafael Wisniewski, Department of Electronic Systems

**ECTS:** 4

**Time:** 26. - 30. June 2017

**Place:**

**Zip code: **

**City: **

**Number of seats: **50

**Deadline: **1. June 2017

**Important information concerning PhD courses** We have over some time experienced problems with no-show for both project and general courses. It has now reached a point where we are forced to take action. Therefore, the Doctoral School has decided to introduce a no-show fee of DKK 5,000 for each course where the student does not show up. Cancellations are accepted no later than 2 weeks before start of the course. Registered illness is of course an acceptable reason for not showing up on those days. Furthermore, all courses open for registration approximately three months before start. This can hopefully also provide new students a chance to register for courses during the year. We look forward to your registrations.

- Teacher: Joachim Dahl
- Teacher: John-Josef Leth
- Teacher: Jan Østergaard
- Teacher: Christoffer Eg Sloth
- Teacher: Rafal Wisniewski

Welcome to Geometric Linear Control Theory

**Description: **In this course, we will study fundamental concepts related to (finite dimensional) linear time-invariant control system such as controllability, observability and stabilizability. These basic concepts will be introduced via the "geometric approach" meaning that they will be related to various subspaces related to the matrices appearing in the system equations. This approach will enable us to introduce the important notion of (A,B)-invariant subspace (and its dual concept, (C,A)-invariant subspace), which will be used to solve the disturbance decoupling problem (and can be used to solve the problem of tracking and regulation). Moreover, the notion of (A,B)-invariant subspace and (C,A)-invariant subspace also turn out to be instrumental in other synthesis problems like observer design, system invertibility, the minimum phase property, and output stabilizability.

**Prerequisites:**

A basic knowledge of mathematics as obtained through undergraduate engineering studies. Knowledge of control is an advantage but not a prerequisite.

Literature: The course will be based on Chapter 4, 5 and 6 of the book [TSH] and the paper [FW]. As preparation it is expected that the participant acquaint themselves with concepts and results corresponding to the material in Chapter 2 and 3 of [TSH].

[TSH] Control theory for linear systems, by Trentelman, Stoorvogel, and Hautus.

[FW] The internal model principle for linear multivariable regulators, by Francis and Wonham.

The book [TSH] can be downloaded from Prof. dr. Harry L. Trentelman's homepage (http://www.math.rug.nl/trentelman/).

**Organizer:**

**Lecturers: **Associate Professor John Leth, Department of Electronic Systems, Associate Professor Mihaly Petreczky, CNRS Lille, France and Professor Rafael Wisniewski, Department of Electronic Systems

**ECTS:** 2

**Time:** 22. - 24. May 2017

**Place:**

**Zip code: **

**City: **

**Number of seats: **50

**Deadline: **1. May 2017

**Important information concerning PhD courses** We have over some time experienced problems with no-show for both project and general courses. It has now reached a point where we are forced to take action. Therefore, the Doctoral School has decided to introduce a no-show fee of DKK 5,000 for each course where the student does not show up. Cancellations are accepted no later than 2 weeks before start of the course. Registered illness is of course an acceptable reason for not showing up on those days. Furthermore, all courses open for registration approximately three months before start. This can hopefully also provide new students a chance to register for courses during the year. We look forward to your registrations.

- Teacher: John-Josef Leth
- Teacher: Rafal Wisniewski

Welcome to Deep Learning

Please see https://phd.moodle.aau.dk/course/view.php?id=791 for course description and registration.