Lakhasly

The study in [218] demonstrates the use of BERT to detect behavioral changes in elderly adults living in smart homes, improving monitoring and personalized care.

The study in [218] demonstrates the use of BERT to detect behavioral changes in elderly adults living in smart homes, improving
monitoring and personalized care. By utilizing BERT’s sequence modeling capabilities, the study identifies anomalies
that may indicate health or behavioral changes. To do this, BERT processes sequences of activities as tokenized events
(e.g., “sleep” or “meal preparation”) and uses its bidirectional attention layers to model typical behavior patterns,
enabling it to flag deviations that might indicate health deterioration. Another study in [254] employs BERT to
enhance indoor positioning accuracy in smart spaces and defend against adversarial attacks. This research presents a
crowdsourced indoor localization system that utilizes Bluetooth Low Energy (BLE) fingerprints. BERT’s self-attention
mechanism captures complex patterns within BLE signal data, distinguishing authentic fingerprints from adversarial
ones. The result is high localization accuracy and improved resilience against database and online attacks, supporting
secure and reliable indoor localization.