Method and apparatus for identifying an object of interest of a specific person based on that person's detectable behavior

A method of identifying an object of interest of a person based on that person's detectable behavior, the method comprising:

• receiving or generating ambient data representing a first time sequence of ambient data items, each ambient data item representing for a specific respective point in time in the first time sequence one or more ambient targets being currently present as a respective object in the spatial environment of the person at that point in time;
• receiving or generating directional sensor data representing a second time sequence of directional data items, each directional data item representing one or more directional measurements relating to a specific point in time in the second time sequence, wherein each of the directional measurements indicates a detected directional aspect of the person's behavior at that point in time;
• matching the directional sensor data to possible targets represented by the ambient data to generate a third time sequence of result data items, each result data item representing a matching result indicating whether a directional data item in the directional sensor data indicates a direction pointing to an ambient target being represented by a temporally corresponding ambient data item in the ambient data;
• and
• classifying one or more of the ambient targets represented by the ambient data based on at least two of the corresponding result data items of the third time sequence to determine which of these ambient targets, if any, is being focused by the person;
• wherein the matching is based on a sensor fusion model configured to determine the most likely focus target in dependence on measurements represented by the first and second time sequences,
• wherein the matching is based on a sensor fusion model comprising a Hidden Markov Model, HMM, which comprises:
• a hidden state space comprising:
• a set of ambient targets that may be focused by the person including the ambient targets represented by the ambient data, and
• an additional state that represents a situation where currently no ambient target is focused by the person;
• and
• an emission space comprising a set of possible directional measurements including the directional measurements represented by the directional sensor data.