Reducing the Cognitive Workload While Operating in Complex Sensory Environments

摘要

A common characteristic of most artificial recognition systems in use today is that they are bottom up approaches, meaning that high-level modules do not influence processing of low-level modules. However, there are some problems and ambiguities at the level of sensory processing, and preprocessing of the signal, that cannot be resolved without taking into account cognitive level expectations. The major goal of our research within this project was to construct a functioning recognition system, based upon fundamental principles of human perception and cognition that exhibits the following properties: The ability to utilize contextual information during the recognition process. The ability to selectively focus attention on important regions of the input data. The ability to adapt preprocessing parameters to changes in the operating environment. The ability to engage in bi-directional information exchange with the user. The system that we constructed implements salient aspects of human perception and cognition such as saccadic eye movements, selective attention and cognitive level feedback during the recognition process, therefore having the potential to reduce the cognitive workload of the soldier when faced with the analysis of complex sensory environments. In this report we provide a brief description of the models and algorithms that we developed within the scope of this project and we present the performance of those models when tested on real- world datasets. We conclude the report with the summary of the most important results.