Investigations on probabilistic analysis synthesis systems using bidirectional HMMs

This contribution deals with preliminary investigations on the behavior of the cortical algorithm in probabilistic hierarchic analysis synthesis systems. Such a subsystem is one the key components of Cognitive Dynamic Systems or Cognitive User Interfaces respectively. Both systems are typically characterized by the cybernetic circle that describes the perception of the environment along the sensory hierarchy, the selection of an optimal response and action articulation on the environment along the motor hierarchy. Further, a cognitive system should be able to predict the consequences of its own actions. For this purpose an inner model of the communication participant and its simulation is required. Based on this assumption, the bidirectional flow of information in analysis synthesis systems may be justified. Even though the cortical algorithm is drawn to cascaded bidirectional HMMs (CBHMMs), in this study the impact of the bidirectional information processing has been investigated just for simple single layer bidirectional HMMs. The proposed experiment is based on Shannon’s channel model, at which synthetic source data are transmitted to the receiver disturbed by Gaussian noise at different SNR. Finally, we compare the state recognition rate for all possible setups using single layer HMMs.