| Fundamentals of Speech Recognition: |
|
|
A Tutorial Based on a Public Domain C++ Toolkit
Introduction Speech recognition systems can be viewed as an application of complex pattern recognition and machine learning algorithms. The development of such a system is a time-consuming and infrastructure-intensive task. The training and evaluation procedures embodied in a state of the art system are loaded with intricate details - most of which seem to make small contributions to performance. The Institute for Signal and Information Processing (ISIP) was founded in 1994 to developed a public domain speech recognition system, and to educate users how to use such a system to achieve state of the art performance. Hence, the primary goal of this tutorial is to facilitate this goal by providing users with detailed step-by-step instructions. This tutorial is divided into into two major parts: explanations of each step in the process which expose the core algorithmic issues, and integrated tutorials which teach you how to build systems. At each step in this tutorial, we have provided detailed examples that you can download and run to verify that your installation is performing correctly. These examples include the necessary data and configuration files to replicate each step in a typical system development cycle. At the end of each section, you will find more extensive examples that provide a more integrated view of the topic presented in that section. Finally, the last section in the tutorial is devoted to examples of how to build some of the most popular recognition applications in use today. Almost every ISIP student and staff member have contributed to the development of the technology presented in this tutorial. For a complete list of our past and present employees, see our personnel pages. If you have questions about the tutorial, please do not hesitate to contact us at ies_help@cavs.msstate.edu. If you have comments about the quality and usefulness of this tutorial, please register your comments in our guestbook. Enjoy, Joe Picone and the Staff at ISIP |
|
|
