Adaptive Sound Zones through machine learning on a large dataset

Level: Master 1
Supervisors: Host Laboratory: Laboratoire d’Informatique de l’Université du Mans (LIUM), Laboratoire d’Acoustique de l’Université du Mans (LAUM)
Location: Le Mans
Beginning of internship: Between January and March 2026
Contact: Théo Mariotte (prénom.nom@univ-lemans.fr)
Application: Send a CV, a covering letter relevant to the proposed subject, possibility of attaching letters of recommendation to Théo Mariotte before December 1st, 2025

Context and objectives

The implementation of differentiated listening zones (sound zones) [1] has applications in many contexts, such as the broadcasting of personalised audio content in vehicle interiors. These methods make it possible to control the acoustic level emitted in defined areas of space, known as bright and dark zones. In the former, the acoustic level is raised to allow the transmission of the useful signal. In the second, the level is attenuated in order to restrict the acoustic signal transmitted to the bright zone. These zones can be created using a network of loudspeakers and microphones.

The methods described in the literature that enable the implementation of differentiated listening zones exploit constrained optimisation (e.g. Acoustic Contrast Control (ACC), Pressure Matching (PM)). These methods do not allow the zones to be adapted to the subject’s movements. Adaptive methods have therefore been developed to overcome this limitation, notably using active control approaches. :

As part of this internship, the aim is to use deep neural networks to adapt sound zones to the subject’s movements. An initial internship enabled models to be implemented in the case of static zones. The objective is therefore to extend existing methods to adaptive cases. Some work has been done on this subject [2] and on the use of neural networks for sound zones [4]. :

Phase 1 :

• Study of the bibliography and familiarisation with the available code base for model learning (Python, PyTorch)
• Improvement of data simulation to simulate small movements of the subject
• Evaluate existing models and classical methods in the context of a subject that can move.

 

Phase 2 :

• Propose new neural approaches to improve reproduction quality when the listening area is moved (see [2])
• Possibility of exploring several architectures and/or formulations of the problem
• Evaluation based on real data (Zhao [3]). Measurements can also be taken.

 
 

Laboratories and supervisory team

The Acoustics Laboratory at the University of Le Mans (LAUM) is renowned for its work in acoustics and has extensive expertise in field reproduction and control methods. Manuel Melon has led and supervised numerous projects on the theme of sound zones.

The Computer Science Laboratory at the University of Le Mans (LIUM) has historically focused on automatic speech processing, with a strong emphasis on deep learning approaches. Marie Tahon works in particular on neural methods for emotion recognition and speech synthesis, with a focus on interpretability. Théo Mariotte works on audio processing methods using neural networks, and in particular develops methods using microphone arrays.

The intern will benefit from the expertise of both laboratories in the areas of acoustics (LAUM) and computer science and machine learning (LIUM).

Candidate profil

Candidates motivated by artificial intelligence and acoustic field reproduction methods, enrolled in a master’s programme in computer science or acoustics.

References

[1]. T. Betlehem, W. Zhang, M. A. Poletti, et T. D. Abhayapala, « Personal Sound Zones: Delivering interface-free audio to multiple listeners », IEEE Signal Process. Mag., vol. 32, no 2, p. 81‑91, mars 2015, doi: 10.1109/MSP.2014.2360707.
[2]. Qiao, Y., & Choueiri, E. (2025). SANN-PSZ: Spatially Adaptive Neural Network for Head-Tracked Personal Sound Zones. IEEE Transactions on Audio, Speech and Language Processing.
[3]. S. Zhao, Q. Zhu, E. Cheng, et I. S. Burnett, « A room impulse response database for multizone sound fieldreproduction (L) », The Journal of the Acoustical Society of America, vol. 152, no 4, p. 2505‑2512, oct. 2022, doi: 10.1121/10.0014958.
[4]. G. Pepe, L. Gabrielli, S. Squartini, L. Cattani, et C. Tripodi, « Deep Learning for Individual Listening Zone », in 2020 IEEE 22nd International Workshop on Multimedia Signal Processing (MMSP), Tampere, Finland: IEEE