Simon listens PIC: 987058033 Grant Agree number: 231143 Project: ASTROMOBILE
15
3.3.1. Parameter definition
In this part we will have to define different parameters which will affect the speech model. This includes
following definitions:
Wordlist: We have to define which words should be recorded for the speech model to cover our
scenarios
Hardware: Which hardware is necessary for recordings, this includes Computers, Microphones,
Soundcards, etc.
Software: With which software should the recording process be done? There are many
possibilities like standard recording software like Audacity or special speech recording solutions
like SSC from the simon speech recognition suite. In our case we will probably use SSC, because it
covers all of our needed functionalities.
People to record: We have to clearly define with whom we record your audio samples. There are
many aspects we should consider to cover your target group like appropriate age range, balanced
ratio of male and female speakers, enough speakers to obtain enough different data etc.
Setting: We clearly have to define which settings should be used. They are for example where the
recordings should happen, under which circumstances etc. Therefor we have to keep in mind,
that the recordings should be as near as possible to our scenarios.
3.3.2. Speech recordings
The second part is the speech recording. In this case we have to consider the conditions defined in part
one. We should also keep in mind that the recordings should be as near as possible to the scenario the
speech model is made for.
3.3.3. Model compilation
The third part is the compilation of the speech data to a speech model. Therefor we will use SAM from
our software suite simon. In this case we will have to make different test in every language we will record
to achieve the best recognition results.
3.4. Recording Italian speech data
In this case we installed a work process to approve the proposed word list and adapt it to the special
requirements of the ASTROMOBILE project and the cultural area of Pisa. So we changed the word list from
111 words to 145 words.
