SIGNAL PROCESS ALGORITHM INTEGRATED DEEP NEURAL NETWORK BASED SPEECH RECOGNITION APPARATUS AND OPTIMIZATION LEARNING METHOD THEREOF

摘要

A deep speech neural network based speech recognition apparatus and its learning method integrated with signal processing algorithms are disclosed. A method of learning a model parameter in a computer-implemented neural network-based speech recognition apparatus according to an embodiment of the present invention includes: (a) a signal processing algorithm for extracting a feature parameter from a speech input signal in a time domain; processing DNN); (b) fusing the deep layer neural network based signal processor with a deep neural network based classifier (classification DNN); And (c) learning model parameters in a deep learning model in which the deep layer neural network based signal processor and the deep layer neural network based classifier are fused.;