University Research software

Carnegie Mellon

Speech Software for Research

The Carnegie Mellon portable speech library (CMPSL) is available for research purposes to universities and government research agencies. The library includes software for recognition and for training, as well as various ancillary programs. We do not have the resources to support the software, so you would be more or less on your own. Some familiarity with current speech recognition algorithms might be necessary to make effective use of this software. The system is programmed in C and has been developed in a Unix environment.

At various times, we have had system components working on the following platforms: NeXT, IBM 6000, H-P 720, Sun SparcStations, DEC Alphas (under Unix) and Pentiums (under Windows NT and 95), so you have a reasonable chance of having it work in your particular environment. Note however that our speech systems require substantial computing resources. For example, to produce acceptable response for a moderate-sized application, you should plan to use a 75MHz Pentium with 40Mb of RAM, or an equivalent workstation configuration. We distribute our software by CD-ROM.

If you're (still) interested, you should send a formal request (on University or Agency letterhead). Your letter should briefly describe the uses you are contemplating for the system. (This helps us in our reporting to our funding agency.) The letter should be addressed to:

Dr. Alexander Rudnicky
School of Computer Science
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh PA 15218

Once we've received your request, we will send you a software license, which you will need to have signed and return. You should return one copy of the license, plus a check for $150 (which helps us cover preparation costs). We will then send the code out to you. Please note that at this time, the CMPSL software is freely available only within the United States. We are working to remedy this situation. Part of our plan is to freely make available binary versions of our system on several major platforms. Look for an announcement of this release in the near future.