Kevin Wilkinghoff



Kevin Wilkinghoff

I am a visiting research scientist at MERL. Before this, I was a research associate at Fraunhofer FKIE while working towards my Ph.D. degree at the University of Bonn, which I obtained in 2024. My research focuses on representation learning for anomalous sound detection, sound event detection and keyword spotting.





News

Ph.D. defense, 02.09.2024
I successfully defended my Ph.D. thesis in Computer Science. My dissertation is available in the EURASIP library.

New position, 15.07.2024
I joined the Speech & Audio Group at MERL as a Visiting Research Scientist.

ICASSP 2023 Outstanding Reviewer Recognition, 18.06.2023
I have been recognized as an Outstanding Reviewer at IEEE ICASSP 2023. In total, 220 out of 5506 reviewers, i.e. approximately 4% of reviewers, have been recognized.

DCASE 2021 Challenge Judge's Award and DCASE 2021 Best Paper Award, 19.11.2021
My submission Utilizing Sub-Cluster AdaCos for Anomalous Sound Detection under Domain Shifted Conditions to task 2 of this year's DCASE Challenge has been awarded with the Judge's award for being the most interesting/innovative method. The proposed approach ranked 3/27. Moreover, the accompanying paper Combining Multiple Distributions based on Sub-Cluster AdaCos for Anomalous Sound Detection under Domain Shifted Conditions I presented at the DCASE Workshop received the Best Paper Award.

Kölner VDI Förderpreis, 18.01.2019
I received a second award for my Master's Thesis Neural Networks in Speaker Identification, namely the Kölner VDI Förderpreis 2018 (third prize). Click here for additional information (in German).

AFCEA Bonn Studienpreis, 30.08.2018
My Master's Thesis Neural Networks in Speaker Identification has been awarded with the AFCEA Bonn Studienpreis 2018 (first prize). More information can be found here (in German).




Full CV: [PDF] (last update: 12/2024) 


Brief CV

2024 - present Visiting Research Scientist
MERL, Cambridge, MA, USA
2022 - 2024 Ph.D. (Dr.rer.nat.) in Computer Science
University of Bonn, Germany
2017 - 2024 Research Associate
Fraunhofer FKIE, Wachtberg, Germany
2014 - 2017 M.Sc. in Computer Science
University of Bonn, Germany
2010 - 2014 B.Sc. in Mathematics
University of Münster, Germany




You can also find me on Google Scholar and dblp.


Preprints

No Class Left Behind: A Closer Look at Class Balancing for Audio Tagging
Janek Ebbers, François G. Germain, Kevin Wilkinghoff, Gordon Wichern, Jonathan Le Roux
Submitted to International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Reference]

Keeping the Balance: Anomaly Score Calculation for Domain Generalization
Kevin Wilkinghoff, Haici Yang, Janek Ebbers, François G. Germain, Gordon Wichern, Jonathan Le Roux
Submitted to International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Reference]


Articles in Peer-Reviewed Journals

2024

Why do Angular Margin Losses Work Well for Semi-Supervised Anomalous Sound Detection?
Kevin Wilkinghoff, Frank Kurth
IEEE/ACM Transactions on Audio, Speech and Language Processing
[Paper][Reference]


Articles in Peer-Reviewed Conference and Workshop Proceedings

2024

AdaProj: Adaptively Scaled Angular Margin Subspace Projections for Anomalous Sound Detection with Auxiliary Classification Tasks
Kevin Wilkinghoff
9th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE)
[Paper][Reference][Code]

Multi-Sample Dynamic Time Warping for Few-Shot Keyword Spotting
Kevin Wilkinghoff, Alessia Cornaggia-Urrigshardt
32nd European Signal Processing Conference (EUSIPCO)
[Paper][Reference]

Analyzing the Impact of HF-Specific Signal Degradation on Automatic Speech Recognition
Fabian Fritz, Alessia Cornaggia-Urrigshardt, Lukas Henneke, Frank Kurth, Kevin Wilkinghoff
International Conference on Military Communications and Information Systems (ICMCIS)
[Paper][Reference]

Strong Label Generation for Preparing Speech Data in Military Applications Using CTC Loss
Fahrettin Gökgöz, Alessia Cornaggia-Urrigshardt, Kevin Wilkinghoff
International Conference on Military Communications and Information Systems (ICMCIS)
[Paper][Reference]

Self-Supervised Learning for Anomalous Sound Detection
Kevin Wilkinghoff
International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Paper][Reference][Code]

F1-EV Score: Measuring the Likelihood of Estimating a Good Decision Threshold for Semi-Supervised Anomaly Detection
Kevin Wilkinghoff, Keisuke Imoto
International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Paper][Reference][Code]

TACos: Learning Temporally Structured Embeddings for Few-Shot Keyword Spotting with Dynamic Time Warping
Kevin Wilkinghoff, Alessia Cornaggia-Urrigshardt
International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Paper][Reference][Dataset]

2023

Language Recognition for SSB modulated HF Radio Signals of Short Duration
Alessia Cornaggia-Urrigshardt, Fabian Fritz, Lukas Henneke, Frank Kurth, Christian Schlich, Kevin Wilkinghoff
15th ITG Conference on Speech Communication (ITG Speech)
[Paper][Reference][Examples]

On Using Pre-Trained Embeddings for Detecting Anomalous Sounds with Limited Training Data
Kevin Wilkinghoff, Fabian Fritz
31st European Signal Processing Conference (EUSIPCO)
[Paper][Reference][Code]

Novel Generative Classifier for Acoustic Events
Paul M. Baggenstoss, Kevin Wilkinghoff
31st European Signal Processing Conference (EUSIPCO)
[Paper][Reference]

Design Choices for Learning Embeddings from Auxiliary Tasks for Domain Generalization in Anomalous Sound Detection
Kevin Wilkinghoff
International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Paper][Reference][Code]

2022

On choosing decision thresholds for anomalous sound detection in machine condition monitoring
Kevin Wilkinghoff, Alessia Cornaggia-Urrigshardt
24th International Congress on Acoustics (ICA)
[Paper][Reference]
 Invited Paper

Towards Human-Machine Integration for Signal Intelligence Applications
Jonas D. Rockbach, Luka-Franziska Bluhm, Isabel Schlangen, Laura Over, Sabine Apfeld, Lukas Henneke, Kevin Wilkinghoff
14th Symposium Sensor Data Fusion (SDF)
[Paper][Reference]

SCALA-Speech: An Interactive System for Finding and Analyzing Speech Content in Audio Data
Alessia Cornaggia-Urrigshardt, Nikita Jarocky, Frank Kurth, Sebastian Urrigshardt, Kevin Wilkinghoff
52. Jahrestagung der Gesellschaft für Informatik (GI-Jahrestagung)
[Paper][Reference]

Speech Recognition Lab
Alessia Cornaggia-Urrigshardt, Fahrettin Gökgöz, Frank Kurth, Hans-Christian Schmitz, Kevin Wilkinghoff
International Conference on Military Communications and Information Systems (ICMCIS)
[Paper][Reference]

2021

Combining Multiple Distributions based on Sub-Cluster AdaCos for Anomalous Sound Detection under Domain Shifted Conditions
Kevin Wilkinghoff
6th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE)
[Paper][Reference]
Best Paper Award

Two-Dimensional Embeddings for Low-Resource Keyword Spotting Based on Dynamic Time Warping
Kevin Wilkinghoff, Alessia Cornaggia-Urrigshardt, Fahrettin Gökgöz
14th ITG Conference on Speech Communication (ITG Speech)
[Paper][Reference]

Sub-Cluster AdaCos: Learning Representations for Anomalous Sound Detection
Kevin Wilkinghoff
International Joint Conference on Neural Networks (IJCNN)
[Paper][Reference][Code]

2020

Using Look, Listen, and Learn Embeddings for Detecting Anomalous Sounds in Machine Condition Monitoring
Kevin Wilkinghoff
5th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE)
[Paper][Reference]

On Open-Set Classification with L3-Net Embeddings for Machine Listening Applications
Kevin Wilkinghoff
28th European Signal Processing Conference (EUSIPCO)
[Paper][Reference]

On Open-Set Speaker Identification with I-Vectors
Kevin Wilkinghoff
Odyssey - The Speaker and Language Recognition Workshop (Odyssey)
[Paper][Reference][Code]

Towards Robust Speech Interfaces for the ISS
Hans-Christian Schmitz, Frank Kurth, Kevin Wilkinghoff, Uwe Müllerschkowski, Christian Karrasch, Volker Schmid
International Conference on Intelligent User Interfaces (IUI) Companion
[Paper][Reference]

2019

Calm Interfaces for Integrated C2 Systems
Hans-Christian Schmitz, Alessia Cornaggia-Urrigshardt, Fahrettin Gökgöz, Samantha Kent, Kevin Wilkinghoff
International Command and Control Research and Technology Symposium (ICCRTS)
[Paper][Reference]

Open-Set Acoustic Scene Classification with Deep Convolutional Autoencoders
Kevin Wilkinghoff, Frank Kurth
4th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE)
[Paper][Reference][Code]

2018

General-Purpose Audio Tagging by Ensembling Convolutional Neural Networks based on Multiple Features
Kevin Wilkinghoff
3rd Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE)
[Paper][Reference]

Accurately Capturing Speech Feature Distributions by Extending Supervectors for Robust Speaker Recognition
Kevin Wilkinghoff
13th ITG Symposium on Speech Communication (ITG Speech)
[Paper][Reference]

Robust Speaker Identification by Fusing Classification Scores with a Neural Network
Kevin Wilkinghoff, Paul M. Baggenstoss, Alessia Cornaggia-Urrigshardt, Frank Kurth
13th ITG Symposium on Speech Communication (ITG Speech)
[Paper]  [Reference]

Robust Detection of Jittered Multiply Repeating Audio Events Using Iterated Time-Warped ACF
Frank Kurth, Kevin Wilkinghoff
International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Paper][Reference]

2017

Glottal Mixture Model (GLOMM) for Speaker Identification on Telephone Channels
Paul M. Baggenstoss, Kevin Wilkinghoff, Frank Kurth
25th European Signal Processing Conference (EUSIPCO)
[Paper][Reference]


Technical Reports for Academic Challenges

2024

FKIE-VUB System for DCASE2024 Challenge Task 2: First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring
Kevin Wilkinghoff, Yacine Bel-Hadj
DCASE2024 Challenge, Ranked 19/28.
[Reference]

2023

Fraunhofer FKIE Submission for Task 2: First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring
Kevin Wilkinghoff
DCASE2023 Challenge, Ranked 4/24
[Paper][Reference]

Fraunhofer FKIE Submission for Task 5: Few-Shot Bioacoustic Event Detection
Kevin Wilkinghoff, Alessia Cornaggia-Urrigshardt
DCASE2023 Challenge, Ranked 6/8
[Paper][Reference]

2022

An Outlier Exposed Anomalous Sound Detection System for Domain Generalization in Machine Condition Monitoring
Kevin Wilkinghoff
DCASE2022 Challenge, Ranked 10/32
[Paper][Reference]

2021

Utilizing Sub-Cluster Adacos for Anomalous Sound Detection under Domain Shifted Conditions
Kevin Wilkinghoff
DCASE2021 Challenge, Ranked 3/27
[Paper][Reference]
Judges' Award

2020

Anomalous Sound Detection with Look, Listen, and Learn Embeddings
Kevin Wilkinghoff
DCASE2020 Challenge, Ranked 8/40
[Paper][Reference]

2019

Open-Set Acoustic Scene Classification with Deep Convolutional Autoencoders
Kevin Wilkinghoff, Frank Kurth
DCASE2019 Challenge, Ranked 19/39 and 3/7
[Paper][Reference]

2018

Open-Set Speaker Recognition with Augmented i-Vectors
Kevin Wilkinghoff
MCE2018 Challenge, Ranked 8/13 and 6/13
[Paper][Reference]

General-Purpose Audio Tagging by Ensembling Convolutional Neural Networks based on Multiple Features
Kevin Wilkinghoff
DCASE2018 Challenge, Ranked 14/39
[Paper][Reference]


Theses

Audio Embeddings for Semi-Supervised Anomalous Sound Detection
Doctoral Thesis, Faculty of Mathematics and Natural Sciences, University of Bonn, 2024
[ULB Bonn entry][EURASIP library entry][Reference]

Neural Networks in Speaker Identification
Master's Thesis, Institute of Computer Science, University of Bonn, 2017
Kölner VDI Förderpreis 2018
AFCEA Bonn Studienpreis 2018

Homogenisierung nichtlinearer elliptischer Differentialgleichungen
Bachelor's Thesis, Mathematical Institute, University of Münster, 2013





Contact

kevin.wilkinghoff@ieee.org