Special Session at Interspeech 2019

Privacy in Speech and Audio Interfaces

An increasing number of devices that feature acoustic interfaces are connected, such as smartphones, televisions, tablets, IoT devices and now also microwave ovens. This trend exposes device owners to threats on privacy. For example, in a recent headline news-story[1], a home-assistant picked up background speech from a neighboring room, triggered the keyword spotter, and transmitted the conversation to a random person. Such behaviour is privacy intrusive, unpredictable and unacceptable. As a speech community, we have to provide solutions that prevent such undesirable behaviour of speech interfaces.

With this special session, we encourage researchers in the speech community to study privacy. Most recent work which combines privacy and speech focused on privacy-preserving approaches which provide the benefits of big-data, while maintaining individual privacy. In addition to continuing in this direction, the special session expands the field and encourages contributions which study the desired level of privacy and explore the nature of privacy; What is privacy? Where do we find it? What characterizes privacy? How do we provide privacy? How should devices behave?

Research in this area is highly interdisciplinary and therefore follows the crossroads-theme of Interspeech 2019. Specifically, we invite submissions from areas such as signal processing, psychology and psychoacoustics, user interface design, cognitive science, machine learning, medical applications, artificial intelligence, law and ethics.

Session structure


Tom Bäckström, Aalto University, Department of Signal Processing and Acoustics, Finland


Stephan Sigg, Aalto University, Department of Communications and Networking, Finland


Rainer Martin, Ruhr-Universität Bochum, Institute of Communication Acoustics, Germany



Tom Bäckström, tom.backstrom@aalto.fi

[1] https://www.theverge.com/2018/5/24/17391898/amazon-alexa-private-conversation-recording-explanation