Privacy preserving distributed speech enhancement in wireless sensor networks

Privacy preserving distributed speech enhancement in wireless sensor networks The last few years have shown a growing interest in distributed noise reduction for speech enhancement in wireless acoustic sensor networks. With this new generation of speech processing algorithms, part of the processing is done at processors not necessarily owned by the user himself. As a consequence, privacy of the users might be at risk. One risk that underlies such distributed speech enhancement systems is the risk of eavesdropping by untrusted parties. Using homomorphic encryption techniques, we have shown that distributed noise reduction algorithms can be turned into privacy preserving algorithms. However, homomorphic encryption leads to a huge increase of the data rate and complexity and becomes impractical. Although it is believed that the future of multimicrophone noise reduction lies in the area of distributed processing, this will only come to practice if privacy can be guaranteed in a practical setting.

Assignment

In this project the goal is to develop a privacy preserving noise reduction algorithm with lower complexity and data rate than the currently existing algorithms.

Requirements

This project requires knowledge on at least:

• signal processing
• stochastic/random processes
• Digital audio and speech processing (in particular speech enhancement)
• experience with Matlab
• cryptography

Contact

dr.ir. Richard Hendriks

Signal Processing Systems Group

Department of Microelectronics

Last modified: 2019-10-21