Accuracy and Robustness in Activity Recognition

Some human activities can be carried out by different humans at different times. The actions that are taking place rather than evaluate how correctly the activity was carried out as well as other qualitative information

High-Level and Long-Term Activity Monitoring

Monitoring high-level activity in large scale dataandreal-world scenario is still difficult to achieve, and it needs proper formalisation of the activities;

Long term activities usually include several sub-activities that may perform in different order

Multi-User and Multi-Sensor Activity Monitoring

In the lab experiments, the data are usually collected by single user activity. However, in the in real life scenario activities can be performed by multiple users concurrently and there may have interaction between them;

Multiple sensors still research challenge

Real World Data Collection

Most of experiments and ADL activities are carried out in laboratories where designs the solutions are based on lab settings. The activities that performed in the lab are also based on the lab environment

Behaviour Trend Profiling and Analysis from monitoring sensor

Long-term monitoring postures some of the challenges such as data labelling and that cause issue of profiling and analysis with data integration

Affective States Detection

How successfully we perform and identify affective states (e.g. happiness, sadness, anger.). Activities in monitoring human physiological parameters could contribute significantly to behaviour trend analysis

Distinguishing between Fall and ADL Events

Distinguishing between fall and the ADL events still poses challenges, though each event has distinct characteristic signatures in the sensor data