A dynamical coherence-detection process has been proposed as primary basis for stable percept creation. This process of coherence detection between neural signals can be realized with plausible neural circuitry, by synchronizing neural oscillators connected through weak synaptic links.
The synchronization process in pools of interconnected neurons calculates a robust estimate from incoming noisy signals and an additional measure of validation. The robust estimate is coded as modulation frequency of the output current of the coherence-detecting layer, and the validation measure is given by the modulation depth of this current. As a worked-out example of the concepts, a stereo vision network was presented, calculating disparity estimates out of real image data by coherence detection.