Speech Recognition Using Morphological Properties and Artificial Neural Networks

这项研究启动奖将允许帕拉卡尔博士试验神经网络方法，从语音频谱中提取说话人不变的特征。最初的工作将对元音和双元音进行分类，并计划将其扩展到塞音、摩擦音、鼻音、液体和滑音。识别结果将与传统方法(如隐马尔可夫模型)的识别结果进行比较。

This Research Initiation Award will allow Dr. Palakal to experiment with neural-network approaches for extracting speaker- invariant features from speech spectra. The initial work will classify vowels and dipthongs, with planned extensions to plosives, fricatives, nasal sounds, liquids, and glides. Recognition results will be compared to those for conventional approaches such as hidden-Markov modeling.