Posts Tagged ‘revision 1090’

Bug in prompts file after folder import

Tuesday, December 29th, 2009

Currently, I have problems to activate simon after importing the backup folder. E.g. simon asked me do add the word “S” to the active vocabulary, but I never had trained this word. Obviously, this is a bug. Take a look at the screen shot:

prompts-bug

simon added an “S” to several words. I don’t know the reason why this happened. But of course that is an explanation why I can’t activate simon any more.

The situation is as follows: my speech model was well-working. I want to restore it, or I won’t continue to test simon (because I don’t want to begin from scratch again and again – I want simon to recognize 1000 different German words; this is a lot of work because I have to adjust the work flow, e.g. fixing problematic wav files with sam). So it is necessary to fix the error that was introduced (I removed a word with sam, then I tried to modify my existing simon model with sam; then the error prevented simon from working).

I have uploaded the backup folder (I posted the download link in the previous post).

I think that the simon import process (simon > Training > Import Trainingsdata > Import Folder) does have the bug that an “S” has been added to a lot of words in the prompts file (you can see that in the screen shot).

It would be nice if this bug would be fixed because it wouldn’t be very user-friendly if you have start from scratch again and again.

By the way, an export function would be fine. Personally, I don’t need an export function because I know where I can find the specific files on my computer. But every good software does have export / import functionality.

I am using Ubuntu 9.10 (64-bit), and the simon svn revision 1090 (plain old trunk).

Segmentation fault

Wednesday, December 23rd, 2009

A few minutes ago, simon crashed. Here is the report. I don’t know yet how to create useful crash reports. I will figure that out later.

At least I can say that simon was able to recognize the word überflutender (SAMPA: y: b ah f l u: t @ n d ah). So it works. I am not sure which sound card is beeing used (I used two microphones at the same time: on-board sound card and USB sound card).

Installing revision 1090

Wednesday, December 23rd, 2009

Today, I made a fresh install of Ubuntu 9.10 (64 bit). Then I did the following steps:

1. $ sudo apt-get install subversion build-essential cmake bison flex gettext gettext-kde kdeartwork kdelibs5-dev portaudio19-dev libxtst-dev libqt4-sql-sqlite libqt4-phonon-dev
2. $ cd 200912
3. $ svn co https://speech2text.svn.sourceforge.net/svnroot/speech2text/trunk simonsource
4. $ cd simonsource
5. $ ./build_ubuntu.sh

Now I have to install HTK:

6. $ tar xvf HTK-3.4.1.tar.gz
7. $ cd htk
8. Removed -m32 from configure.ac with gedit. This step is necessary because I am using a 64-bit system.
9. $ sudo apt-get install autoconf
10. ubuntu-64bit@ubuntu-64bit-desktop:~/200912/htk$ autoconf
11. $ ./configure
12. $ make
13. $ sudo make install

Now I will have to continue with the configuration.