<ul><li>ADLGen is a generative framework designed to synthesize realistic sensor sequences for human activity modeling in ambient assistive environments.</li><li>It integrates a decoder only Transformer with symbolic temporal encoding and a context-aware sampling mechanism to generate semantically rich sensor event sequences.</li><li>ADLGen incorporates a large language model to verify coherence and generate correction rules automatically, enhancing semantic fidelity and correcting structural inconsistencies.</li><li>Experimental results show that ADLGen outperforms baseline generators in statistical fidelity, semantic richness, and downstream activity recognition, providing a scalable and privacy-preserving solution for ADL data synthesis.</li></ul>

ADLGen: Synthesizing Symbolic, Event-Triggered Sensor Sequences for Human Activity Modeling

Discover more