Abstract

This work explores the trade-offs between time and frequency information during the feature extraction process of an automatic speech recognition (ASR) system using wavelet transform (WT) features instead of Mel-frequency cepstral coefficients (MFCCs) and the benefits of combining the WTs and the MFCCs as inputs to an ASR system. A virtual machine from the Speech Recognition Virtual Kitchen resource (www.speechkitchen.org) is used as the context for implementing a wavelet signal processing module in a speech recognition system. Contributions include a comparison of MFCCs and WT features on small and large vocabulary tasks, application of combined MFCC and WT features on a noisy environment task, and the implementation of an expanded signal processing module in an existing recognition system. The updated virtual machine, which allows straightforward comparisons of signal processing approaches, is available for research and education purposes.

Advisor

Rebecca Bates

First Committee Member

Vincent Winstead

Second Committee Member

Qun (Vincent) Zhang

Date of Degree

2016

Language

english

Document Type

Thesis

Degree

Master of Science (MS)

Department

Electrical and Computer Engineering and Technology

College

Science, Engineering and Technology

Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Share

COinS