Master 2 Research Internship – Human Activity Recognition, Dataset Synthesis and Analysis Methods

Job Description

IMT Atlantique and IMT Nord-Europe are leading Technological University within IMT (Institut Mines-Télécom)—France’s largest group of engineering schools. Their mission is to support societal transitions, train responsible engineers, and harness scientific and technical excellence for education, research, and innovation.

The position is based in Brest, within the RAMBO team (Robot interaction, Ambient system, Machine learning, Behavior, Optimization – IMT Atlantique) at the Lab-STICC laboratory (UMR CNRS 6285). This research project is conducted in collaboration with the HIDE team (Human Interaction Data Ecosystem – IMT Nord Europe) at CERI SN. The teams share a common interest in analyzing interactions (between humans and systems), with a particular focus on applications involving assistance for vulnerable individuals. The ObsAVQ project aims to develop methods for generating reports on Activities of Daily Living (ADL) performed at home over extended observation periods.

Scientific Context and Internship Objectives

In the context of monitoring health status changes at home, the focus is on generating reports regarding Activities of Daily Living (ADL) [1] observed in the home environment, intended to prepare for pre-consultation discussions with a primary care physician. The project assumes the availability of a list of events—characterized by semantic annotations (recognized activity type) and temporal data (timestamps)—identified by sensors integrated into the home; the objective is to identify trends over long durations, ranging from one week to several years.

More specifically, within the wide range of ADLs, the work focuses on: (1) sleep, (2) nutrition, and (3) social interactions. Data regarding these activities will be derived from an acoustic analysis of the home environment.

Algorithms for analyzing Activities of Daily Living (ADLs) require labeled training data (ground-truth datasets). Compiling a dataset using real users over the long term is costly and time-consuming. Consequently, this study focuses on dataset “synthesis” techniques [2]: the approach involves first developing a behavioral model, then running it over the required simulation period while varying parameters to challenge the learning and classification algorithms.

Regarding the methodology for dataset construction, we will use medical descriptions of symptoms observed during health anomalies—specifically their type, frequency, and time of occurrence. The goal of the machine learning process will be to identify the anomalies that caused the symptoms exhibited by the individual.

Objectives:

* O1: Generation of a home-based ADL activity log dataset derived from behavioral models, with ground truth known by design. * O2: Development of an offline method for automatically generating activity reports using machine learning, trained on the dataset generated in O1. * O3: Evaluation of the proposed method using the dataset generated in O1, and comparison with the state of the art.

References:

1. Edemekong, Peter F., Deb Bomgaars, Sukesh Sukumaran, and Shoshana B. Levy. “Activities of daily living.” (2019). 2. Weihang You, Hanqi Jiang, Jiahui Li, Zishuai Liu, Tianming Liu, Jin Lu, and Fei Dou. 2026. ADLGen: Synthesizing Symbolic, Event-Triggered Sensor Sequences for Smart-Home Human Activity Modeling. In Proceedings of the 2026 ACM/IEEE International Conference on Embedded Artificial Intelligence and Sensing Systems (SenSys’26). Association for Computing Machinery, New York, NY, USA, 1029–1042.

Position Requirements

Skills and Knowledge:

* Candidates must be currently pursuing a Master 2 degree in Computer Science (5 years of higher education or equivalent). * Programming and machine learning skills are highly desirable (demonstrated via internships, reports, publications, code samples, etc.). * Initial research experience or basic training. * Proficiency in French and English (reading, writing, and speaking).

Conditions

* The internship is scheduled for a duration of 6 months, starting in January/February 2027. * Application deadline: 01/12/2026. * An interview lasting approximately 30 minutes (video conferencing available) to discuss your background, motivation, skills, and career goals. * Stipend: 4.50€/h (i.e about 580-640€/month) * Location: Plouzané (29), France – i.e. the Brest campus of IMT Atlantique

Application Package Requirements

* Detailed CV. * Cover letter explaining your interest in the topic and your relevant skills. * Academic transcripts for M1 and M2 (or equivalent).

Contacts

* Christophe LOHR, Associate Professor, christophe.lohr[at]imt-atlantique.fr * Mihai ANDRIES, Associate Professor, mihai.andries[at]imt-atlantique.fr * Anthony FLEURY, Full Professor, anthony.fleury[at]imt-nord-europe.fr