INTEREST DETECTING METHOD AND APPARATUS BASED FEATURE DATA OF VOICE SIGNAL USING DEEP BELIEF NETWORK, RECORDING MEDIUM RECORDING PROGRAM OF THE METHOD

摘要

The present invention relates to a method for detecting preference based on voice by using a deep belief network (hereinafter, referred to as DBN), which comprises the following steps: receiving learning data composed of voice signals and preference identification information corresponding thereto to extract feature information on voice signals, learning the feature information through the DBN to generate DBN data, and adding a label layer composed of preference identification information to the DBN data to complete generation of the DBN data; and extracting feature information from test voice signals when the test voice signals are inputted to detect DBN data corresponding to the feature information and output a result of the detected preference according to the preference identification information recorded on the label layer of the detected DBN data.