Project details

Multi-mic array signal generation

Published by: Major company in signal audio processing

Status: READY FOR PUBLISHING

Category: SIMULATION & OPTIMIZATION

Application domain: OTHER

Budget (EUR): UP TO 15000

Project description

In the context of the development of next generation multi-microphone algorithms, the project consists of creating a simulator for automatically generating multi-microphone array recordings starting from a speech signal and a noise signal.



According to our experience, a simple model using the room impulse response is not realistic enough, so we're looking for a more advanced model.



The acoustic scenario to be simulated is that of a microphone array picking up the speech of a talker, along with environmental noise.



Acoustic parameters of the model:

- Indoor or outdoor, so with or without room reflections. Some basic control of reverberation time and diffuseness should also be available.

- The type of microphone array (2 to 7 omnidirectional microphones). The array can be considered in the center of the room, where applicable.

- Distance between the talker and the microphone array

- Types of noises:

1. diffuse noise

2. non-diffuse noise


The input to the project will be:

- exemplary N-mic recordings captured with N-mic arrays.

- input noise files (mono)

- input speech files (mono)

Project goal

The project goal is to:

- develop a framework in either python or matlab, that will allow the generation of N-mic recordings.

- develop a test framework to validate how close the generated recordings are to real N-mic recordings. The quality of the proposed validation method is crucial for the selection of the solution.