Problem Set 6
From MoDe
| Table of contents |
Speech Recognition
Introductory Stuff
sample code for speech recognition
Problem Assignment
In problem set 5, content was associated with locations. There might be many locations and it becomes difficult to do selection from a long list of items. It is much easier to say the place.
Assoicate the <object> with <place> Remove <object> from <place> Show me the <objects> associated with <place>
I am now at <place> Call this place <place> This is the place that I call <place>
<object> is picture, video, text note, audio note
and <place> is home, office, supermarket, hardware store, toy store, computer store,
or any other place you can think of.
6.883 Speech Server
We have our own speech server set up at http://mini-ozone.csail.mit.edu:9000/RPC2. To help you debug, you can use the following simple sample client program (http://web.mit.edu/6.883/files/sls-sample-client.tar.gz) that sends a raw audio file or a wave file repeatedly.
Additionally, you can also perform the speech processing using your IPAQ. There is a small utility program - gpe-soundbite, that can record and play recordings. There is no way to access the functionality directly, but you can invoke the program using python's os shell interface. Unfortunately, I don't know what file format the recordings are saved as (seems like a compressed wave format), so I believe they need to be converted somehow before they can be used. If you discover more or know a solution, please post it below.
