自动语音识别研究的人工神经网络方法

ARTIFICIAL NEURAL NETWORKS APPROACH TO AUTOMATIC SPEECH RECOGNITION

本文概括介绍支撑自动语音识别(ASR)技术的科学原理,简要叙述工程技术、计算机技术和数学对发展这一技术的贡献。然后围绕自动语音识别三大基本原理,讨论人工神经网络(ANN)在这一领域研究的典型问题、方法原理、求解方法、性能水平对比,以及局限性和进一步研究的问题。

In this paper, the science behind automatic speech recognition(ASR) technology is outlined and the contributions of engineering, computer science, and mathematics to this field are briefly described. Artificial Neural networks(ANN) approach to ASR is then discussed in terms of three principles forming the foundation of a primitive automatic speech recognition technology, with emphases on typical questions, methodologies, question-solving strategies, performance comparisons, constrains and further investigations of this newly emerged approach of ASR.