neural network and application thereof for speech recognition.

摘要

Neural network system comprising a plurality of levels (20, 21, 22) of interconnected neural networks (10, 11), each of said neural networks interconnected temporary neural network space (1), said levels input level (20) which can apply a sequence of input vectors (2) with at least a first rate, in which system at least two levels (20, 21, 22) are interconnected by at least one decimator ([102]) for reduce the rate of the signal supplied or at least one interpolator ([103]) for increasing the rate of the supplied signal.