{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}


Assignment_5 - Massachusetts Institute of Technology...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Massachusetts Institute of Technology 6.345 Automatic Speech Recognition Spring, 2005 Issued: 4/22/05 Due: 5/11/05 Assignment 5 Automatic Speech Recognition Systems This lab is intended to familiarize you with the SUMMIT recognizer which is a landmark and segmental-based recognition system. You will train and test a complete speech recognition system using the Pegasus corpus which consists of flight status inquiries. Proceeding step by step through the entire process, you will examine the information/data that is needed at each stage, build/train the necessary models, and evaluate the performance and behavior of the models. In the process, you will become familiar with the components of the SUMMIT speech recognition system. Another type of recognizers that you are given the option to explore is the HMM-based SPHINX. SPHINX 3 was originally developed at CMU and SPHINX 4 is an open-source state- of-the-art system written in JAVA. Since we feel that the SPHINX lab in its current setup is less developed, we have decided to include it as an optional section. Students who wish to explore this type of speech recognition systems for their research or for pure interest may find the SPHINX lab useful and are encouraged to do so. Any feedback or suggestions concerning this lab are welcome. Please read the appendices carefully for all information pertaining to the files and commands used in the lab before starting. 1 An Introduction to the SUMMIT Speech Recognition System 1.1 Getting Started To get started with the first part of the lab, type the following command at the UNIX prompt: % start_lab5.cmd This will create a new subdirectory under /s/6.345/users/$USER called lab5 and pop- ulate it with the set of files needed for this lab. These files contain data, models, and parameter specifications used during the lab. Descriptions of the files can be found in the SUMMIT Files appendix at the end of this handout. In the first part of the lab, we will look at some of these files in detail. 1
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
1.2 Data Files In this part of the lab, we will examine the different data files needed to build a speech recognition system. In particular, we will look at the following: the partitioning of a corpus into disjoint training, development, and testing sets, audio waveforms, word and phonetic level transcriptions, and pronunciation dictionaries/lexicons. 1.2.1 Corpus The SUMMIT system makes use of a corpus file which contains specifications of all the data for several domains. The corpus file has already been created and can be found in the file galaxy.corpus . The program corpus tool can be used to examine the properties and sets defined in the corpus file. T1:
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page1 / 37

Assignment_5 - Massachusetts Institute of Technology...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon bookmark
Ask a homework question - tutors are online