Automatic Prediction of Olfactory Recognition Events in Detection Dogs Using Wearable Transducers

Principle Investigator: Dr Firat Guder
Research Group: Guder Research Group
Funding Agency: Institute for Security Science and Technology

Project Summary:

In this project, our eventual aim is to solve a practical problem; that is, improving the accuracy of detection dogs in identifying target odours and increasing the efficiency of training of dogs (thus reducing costs).

There is strong evidence in the literature that the patterns of respiration/sniffing and heart rate vary significantly in dogs, when performing tasks that are associated with a reward. In addition, the speed at which a dog moves or the amount of time it spends at the source of an odour during an odour discrimination task also provides information about olfactory recognition events.

We hypothesise that, combining these three modalities -- i.e. the patterns of respiration/sniffing, heart rate and motion) will yield enough information to accurately identify an olfactory recognition event in sniffer dogs using machine learning approaches. We will develop a non-invasive wearable transducer system that can measure the patterns of respiration/sniffing, heart rate and patterns of motion of a dog
and predict automatically when an olfactory recognition event occurs during an odour discrimination task.