You are here: Experiments / Virtual Microphone

Virtual Microphone Experiments

The aim of this work is to capture the signals of a remote distant speaker using discreet arrays of static microphones.  This is achieved through complex signal processing.

The functional design of the signal processing system capable of developing a “virtual” microphone using signals from an array of microphones has been built and tested in lab conditions.  A recording of a man and a woman speaking together in the same room (not to each other just speaking at the same time) was played through loudspeakers.  Attempts to generate virtual microphones have been attempted using a variety of internal parameter settings.  The system assesses diffuseness and direction of the audio captured from the two arrays of microphones used to capture the sound.  The signal processing then attempts to reconstitute an audio signal from a given location in the sound scene.

  

System diagram for the Virtual Microphone Signal Processor

Photograph of lab set-up used for testing the virtual microphone performance

Male / Female


Recording is the input from a single microphone as shown.

Female / Male


Recording is the input from a virtual mcrophone located near to the female voice.

Female / Male


Recording is the input from a virtual microphone near to the male voice.

Exploitation/Further work

The initial results are encouraging but more work is required to offer better isolation of the two virtual microphone signals.  Efforts will be made to improve beam forming, which effectively narrows the arc from which the virtual microphone records signals and audio zoom which places the microphone closer to a particular location in the sound stage.

  • System integration is required to make the code run faster and to work with other bodies of code.  A active MQ API needs to be defined and implemented and GUI developed to help control the placements and performance of the virtual microphone in the sound scene.
    • The concept is ideally suited for the performance use case – though could also be used (in principle) to capture audio within a classroom without needing to close mic the teacher.

We also want to confirm using listening test that the measured improvements are mirrored by the perception of people.  We plan 3 listening experiments:

  • Exp.1: Test whether listeners perceive a change in the distance from the source to different VM positions in a similar way as it would be perceived when using real microphones placed at the same distances as for the VM.
    • Exp.2: Test whether the spatial image captured using virtual stereo microphone recording techniques is perceived by listeners in a similar way as it would be perceived when the spatial image  was captured using real stereo recording techniques.
    • Exp.3: Test how close perceptually are the generated VM signals to the real microphone signals for the same microphone type, position and orientation of the spatial recording (stereo recording using 2 cardioid microphones – both real and virtual).

 

 

We will watch progress but we do not expect to be using the microphone in either use case as the code base and technology will be too experimental.

 

Developments and signal processing approaches may form the subject of patents and may be used in implementations of audio capture offered by Fraunhofer into standards bodies or in Fraunhofer audio software.