NavigationSearch |
ASR Systems for teachingSubmitted by mapb on Tue, 2006-12-05 10:01.
Hi, does anyone have any experience in using Sonic, Sphinx, HTK or some other "free" system for teaching purposes in undergraduate courses? What is better? How did it go? Ideas? ( categories: Automatic speech recognition )
I used Sonic since I was working at CSLR, Colorado (where Sonic was developed). I would say you won't get the updates (new algorithms implemented) that frequently, and I don't think that you will find many people using it. There is a question mark how much interest the developers have for making it still publicly available since they already left CSLR. For long-term usage, I would go with either HTK or Sphinx. For your purpose, it will be a good choice since it will be easier for a first-time user to get used to it (simple manual, sample scripts to run experiments, etc.). -Murat I just got a non-commercial licence for Sonic (check CSLR website) and i'm going to try it out. At a first glance it seems really fine as a starting point, as Murat pointed out, since it comes along with a good tutorial on digits and some examples. Moreover, it has already been ported to several languages. I'll report my experience with it. This is a TEST Comment I have a preliminary business plan that requires a specialist in the field of automatic speech recognition. I am a speech layman, however would like to get an expert opinion for an initial feasibility study. I am in the New York area, and would happily pay for dinner/transport etc. Anyone with an entrepreneurial flair and who would be available to answer some simple questions, please get in touch! Thanks, James |
Courses |
Hi,
Having enjoyed my own HTK-based recognition tutorials as a student, I designed HTK-based speech recognition tutorials and smaller HTK-based tasks/demonstrations for 3rd year bachelor students. Being a regular user of HTK, I think it is a good toolkit to demonstrate the structure and interplay of components of standard speech recognisers in a quick and easy way.
Because none of the students had any programming skills, because most of them even lacked experience with unix/linux and because I wanted the students to focus on speech recognition rather than on technical practicalities, the tutorials were designed to get the students through the practicalities of HTK as smoothly as possible. This worked really well. The students judged the tutorials and the smaller tasks/demonstrations to be useful tools that helped them understand the theoretical parts of the course better.
I didn't use the other systems in any of my courses yet.
Hope this helps,
Christophe.