ニューラルネットワークを用いた音色に対応する感性モデル

概要

論文の詳細を見る
This paper presents a "Kansei" model for musical timbre using a neural network. The three-layered network, which is constructed based on the structure of auditory sense, consists of seven input nodes, four neurons in the hidden layer, and three neurons as the output nodes. The input is a pattern of tone pressure levels for harmonics and the output is a three-dimensional vector of the Kansei information that is characterized by "clearness-unclearness", "thickness-thinness" and "sharpness-softness". After training the network, the analysis of the synaptic weights of all the neurons gives the following results that agree well with the knowledge of psychology and physiology. 1) Each harmonic equally influences the "thickness-thinness". 2) The higher harmonics give a contrary effect to the "clearness-unclearness". 3) Two neurons with lateral inhibition appear in the hidden layer and extract the 6 th and 7 th harmonics. Another neuron in the hidden layer extracts the 5th, 6th and 7th harmonics. These three neurons determine the "sharpness-softness".
社団法人日本機械学会の論文
2000-12-25