Adversarial Deep Learning on Speech-To-Text
|Typ||Master, Forschungspraxis, IDP|
|Betreuer||Dipl.-Ing. (Univ.) Ludwig Kürzinger
Tel.: +49 (0)89 289-28562
|Sachgebiet||Speech Recognition, Deep Learning|
|Beschreibung||Speech Recognition enables a machine to understand human voice and convert it to text. State-ofthe-art methods such as DeepSpeech are able to increase accuracy using end-to-end deep learning .
However, a recent publication describes a method how to deceive DeepSpeech to recognize a different sentence instead of the spoken one . Improving the robustness of deep-learning–based speech recognition against such attacks remains a challenge and will be the topic of this thesis.
The main tasks for this topic will be to get familiar with the Mozilla DeepSpeech framework, and then use the provided (english/german/mandarin) dataset to train your own model.
In the second stage of the topic, the task is to train an adversarial model and to propose countermeasures against such attacks.
 2014, Deep Speech: Scaling up end-to-end speech recognition
 2017, Audio Adversarial Examples: Targeted Attacks on Speech-to-Text
|Voraussetzung||- Experience with Python and/or C++
- Interest in machine learning
- Independent work style
- Motivation to learn new concepts
|Bewerbung||If you are interested in this topic, we welcome the applications via the email address above. Please set the email subject to "<Type of application> application for topic 'XYZ'", ex. "Master’s thesis application for topic 'XYZ'", while clearly specifying why are you interested in the topic in the text of the message. Also make sure to attach your most recent CV (if you have one) and grade report.|