Speech Recognition in noisy environment using Deep Learning Neural Network
Year: 2017
Type: dissertation
Abstract: Recent researches in the field of automatic speaker recognition have shown that methods based \non deep learning neural networks provide better performance than other statistical classifiers. On \nthe other hand, these methods usually require adjustment of a significant number of parameters. \nThe goal of this thesis is to show that selecting appropriate value of parameters can significantly \nimprove speaker recognition performance of methods based on deep learning neural networks. \nThe reported study introduces an approach to automatic speaker recognition based on deep \nneural networks and the stochastic gradient descent algorithm. It particularly focuses on three \nparameters of the stochastic gradient descent algorithm: the learning rate, and the hidden and \ninput layer dropout rates. Additional attention was devoted to the research question of speaker \nrecognition under noisy conditions. \nThus, two experiments were conducted in the scope of this thesis. The first experiment was \nintended to demonstrate that the optimization of the observed parameters of the stochastic \ngradient descent algorithm can improve speaker recognition performance under no presence of \nnoise. This experiment was conducted in two phases. In the first phase, the recognition rate is \nobserved when the hidden layer dropout rate and the learning rate are varied, while the input \nlayer dropout rate was constant. In the second phase of this experiment, the recognition rate is \nobserved when the input layers dropout rate and learning rate are varied, while the hidden layer \ndropout rate was constant. The second experiment was intended to show that the optimization of \nthe observed parameters of the stochastic gradient descent algorithm can improve speaker \nrecognition performance even under noisy conditions. Thus, different noise levels were \nartificially applied on the original speech signal.
Author: A. A. Nasef
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3083659827