COMBINED LEARNING METHOD AND APPARATUS USING DEEPENING NEURAL NETWORK BASED FEATURE ENHANCEMENT AND MODIFIED LOSS FUNCTION FOR SPEAKER RECOGNITION ROBUST TO NOISY ENVIRONMENTS

摘要

A method and apparatus for combined learning using a deep neural network-based feature enhancement and modified loss function for robust speaker recognition in a noisy environment are presented. According to an embodiment, a method for reinforcing features based on a deep neural network and combining learning using a modified loss function includes: learning a feature reinforcement model based on a deep neural network; Learning a speaker feature vector extraction model based on a deep neural network; Connecting the output layer of the feature enhancement model and the input layer of the speaker feature vector extraction model to each other; And performing joint learning in which the connected feature reinforcement model and the speaker feature vector extraction model are regarded as one model and additionally learned.