DEVICE FOR SPEECH DEREVERBERATION BASED ON WEIGHTED PREDICTION ERROR USING VIRTUAL ACOUSTIC CHANNEL EXPANSION BASED ON DEEP NEURAL NETWORKS

摘要

A WPE-based reverberation canceling apparatus using a virtual channel extension based on a deep neural network according to an embodiment of the present invention includes: a signal receiving unit for receiving a first voice signal through a single channel microphone; a signal generator for generating a second voice signal by applying a virtual channel extension algorithm based on a deep neural network to the first voice signal; and applying a deep neural network-based two-channel weighted prediction error (WPE) algorithm to the first and second speech signals to remove the reverberation of the first speech signal and generate a dereverberation signal from which the reverberation is removed. It may include a reverberation removing unit for