Week 5
The big thing happened this week was Quarter walkarounds. ETC faculties came to check our progress and provided many useful feedbacks and advices.
For the current design of our prototype, we have a guide character that is used for providing tutorial and encouragement, feedbacks and other stuffs in the game when the kids are supposed to read. We are thinking about re-designing it into a lower status listener that kids are willing to read to. And we are also considering other roles and actions suitable for this character and other potential characters as well.
Two biggest challenges we have for this project for now are setting up playtest with children and getting the voice recognition working. For the playtest issue, we have scheduled a time at Children’s Museum of Pittsburgh. And we are also trying to contact faculties who have young children and do some testing and exploration with them. If you have a young child at home and willing to help this project, please contact us!
As for the voice recognition, we have successfully gotten a small demo running. Basically we are using CMU Sphinx framework. The main reason for choosing it instead of Google Speech API is because it is highly-customizable. It allows us to define how a word is pronounce by setting its phonemes, the grammars of sentences and it is possible to use a dedicated dataset that is constructed based on children’s voice instead of a general dataset. And it also supports keyword spotting and can produce realtime analysis results by having it running locally. In our prototype, we are using PocketSphinx 5 prealpha, and use emscripten to compile it into javaScript and have it running on our HTML5 controller. For now, the performance seems not bad and we are going to have further testing and development on it.