Signal processor and learning method thereof

Abstract

Abstract A signal processor which executes formula operations at a high speed for input signals such as image information. The weight coefficients corresponding to a DCT (discrete cosine transformation) are set for a neural network NH of an n-layer structure which comprises an input layer I1, (an intermediate layer H1), and an output layer O1. The input layer I1 comprises the same number of interfaces as the number (64 in this case) of pixels included in a block (12), and the output layer O1 comprises the same number of neurons as the number of DCT coefficients to be outputted. When the image information of the block (12) is inputted to the neural network NN, the DCT coefficients a(0)-a(63) which are the result of the DCT processing of this image information are outputted immediately. The weight coefficients are determined by making the neural network learn the DCT coefficients as teacher data obtained through actually operating the same picture information as that inputted into the neural network NN using formulae in a DCT transformation section (18).

Inventors

• 👤 Ryoichi Technical Research Division Dangi

Bibliographic Details