Activity Recognition and Order-Based Anomaly Detection for Smart Homes

EasyChair Preprint 14911

Abstract

Sensors provide an unobtrusive way of collecting daily living activities data of users. With the advancement in the field of electronics, collection of sensor data has become easier and can be used to create smart systems to assist users. This paper presents a deep learning approach with two stages: activity recognition and anomaly detection. Different LSTM models are studied for activity recognition and recognized activities are used to detect the abnormal behaviour of the user based on the sequences generated by PrefixSpan algorithm. The performance of the approach has been evaluated on real smart home dataset collected by CASAS on the Aruba testbed for the duration of 8 months.

Keyphrases: Smart Home Activities, activity recognition, anomaly detection, frequent sequential patterns

@booklet{EasyChair:14911,
  author    = {Jinal Patel and Nalinadevi Kadiresan},
  title     = {Activity Recognition and Order-Based Anomaly Detection for Smart Homes},
  howpublished = {EasyChair Preprint 14911},
  year      = {EasyChair, 2024}}