Post-doc position for specialist in spatial audio signal processing and machine learning

Short term contract

1 PostDoc position for specialist in 3D audio signal processing and machine learning

Duration : 12 months

Availability : as soon as possible, not later than november 2022


IRCAM is a non-profit association, associated with the Centre National d'Art et de Culture Georges Pompidou and located in Paris at 1, place Igor Stravinsky, 75004 PARIS. Its missions include research, creation and pedagogical activities on 20th century music and its relationship with science and technology.


Within the framework of the HAIKUS research project financed by the ANR (French National Research Agency), IRCAM is recruiting a Research and Development Officer on a 12-month fixed-term contract. The objective of the HAIKUS project is the joint exploitation of machine learning (ML) and audio signal processing methods to solve acoustic problems encountered in audio augmented (AA) reality applications. In particular, the AA methods are applied to automatically identify the acoustic channel between one or more sources and the listener. The three main objectives of the project are (a) the blind estimation of the acoustic cues of the room and/or its geometry from the observation of the reverberated audio signals emanating from the sources present in the room, (b) the inference of rules for modifying the spatialization  parameters or interpolating room impulse responses according to the listener's movements, and (c) the blind estimation of the listener's HRTFs from the binaural signals captured in the environment thanks to in-ear microphones. In the framework of this project, IRCAM is more particularly interested in this last objective, but the links with the two previous questions are naturally important.

The project is coordinated by IRCAM and brings together the LORIA in Nancy and the MPIA laboratory of Sorbonne University. These three laboratories share expertise in audio signal processing, machine learning, acoustic imaging and modeling, and binaural technologies.

The project takes place within the Acoustic and Cognitive Spaces team (EAC), integrated into the UMR STMS IRCAM/CNRS/Sorbonne University/Ministry of Culture. The EAC team (Acoustic and Cognitive Spaces) is particularly interested in the analysis/synthesis and perception of immersive sound scenes.


  • Automatic selection of HRTFs (Head Related Transfer Functions) among public HRTF databases from binaural recordings in unsupervised conditions (non anechoïc conditions, unknown signals, moving sources, moving listener).
  • Tests and improvement of DNN algorithms developed in the team applied to simple real-world use cases (single source, moderate and diffuse reverberation).
  • Extension of the DNN algorithms for automatic HRTF estimation, i.e. without relying on measured HRTFs databases.
  • Extension of previous methods and algorithms in order to take into account environments with significant first reflections.
  • Contribution to the development/recording of training and test databases.


  • The candidate should have a PhD degree with a strong component in machine learning and audio signal processing.
  • Extensive knowledge of deep learning algorithms and experience with their application to the audio domain.
  • Excellent knowledge of Python and Tensorflow.
  • Good knowledge of Linux and Mac OS X.
  • Excellent level of English.
  • Experience in writing peer-reviewed journal/conference papers.

SALARY : According to background and experience.


Please send a cover letter with reference 2022_HAIKUS, a CV detailing the level of experience/expertise in the above-mentioned areas (and any other relevant information) and letters of recommendation to