Automatic Gesture and Movement Recognition for Child Behavioural Analysis: A Systematic Review of the Laboratory-to-Natural Setting Gap

Automatic gesture and movement recognition techniques are mainly used with adults for various purposes in public, clinical, and laboratory settings. Growing interest in this field has led to the increasing application of these methods in child behavioural analysis to serve different societal and educational functions. However, manual human annotation of behaviours remains the predominant method, and only a limited number of studies have explored the use of automatic recognition for children. This review aims to evaluate the rapidly developing techniques of automatic gesture and movement recognition that focus on child behaviour analysis across different settings and for different purposes. More specifically, it analyzes their purposes, target groups, settings, accuracy, and limitations, as well as the ethical issues and data privacy frameworks that should be considered in child-centred AI. Using a systematic review approach following the PRISMA guidelines, this study examines research published between 2021 and 2025 in four databases: Web of Science (WoS), Scopus, PubMed, and IEEE Xplore. From a total of 27 included studies, the findings reveal that automatic gesture and movement recognition is being applied across multiple fields, with consideration of children’s developmental needs. However, a critical gap in technical reporting was identified: fewer than half of the included studies (44%) provided accuracy metrics or clinical validity. Furthermore, evidence of robust ethical safeguards remains limited. To support children’s well-being, future studies must bridge the lab-to-field gap, prioritize natural research settings and enforce ethical and data protection measures.