We are getting ready for our final presentation. It seems we can show two versions of our project. We have also finished filming and will present the results to our test audience. The slides will be posted after the presentation.
This week we will show a first pre-alpha prototype with rudimentary functions for both modalities.
There were as usual problems which slowed our progress considerably. The kinect – we are currently working on – do not support Windows on a VM. We have however managed to find solution which we are implementing right now. As far as we see we wont have a full functioning gesture detection for now.
As for our final presentation the video production team has already start filming.
Since last week we have been discussing to switch to C#. We found a solution for Mac Users to use C# so that in the end we decided to focus on the add-in for MS Powerpoint with C#. Both part of the functionality are working although there are some minor issues to solve. We will now finalize our work and fuse them into one add-in. This will be the last part of our prototyping phase which will then be thoroughly tested. The design is already in place.
As for our media task we have decided to push the filming back for another week.
For our prototype we have developed an add-in for MS Powerpoint which will be the interface for the user.
As previously mentioned we have decided to use Java. It turns out that in order to use the SDK for the Kinect which requires C#/C++ we have to use the a plug-in (Jnect). It creates a interoperability between the MS SDK and Java.
So far our problems is to connect our interface to the Java code. It seems that MS Powerpoint only supports C#/C++ for add-ins. We are currently considering to rewrite our code in C# because of their similar nature.
For those who missed our presentation last week you can find it here.
Here is a short update of our progress.
Given our diverse field of studies we have divided the task of our project in 3 parts. So that the three following tasks can be worked on simultaneously.
1. Implement gesture recognition of the Kinect and add gesture command to the application.
2. Implement speech recognition and add voice command.
3. Drafting our storyboard for our final video presentation. And make necessary preparations for the principal photography.
Due to some unforeseen problems there is not much progress on the first part of our project. We hope to have a prototype in the following week. As for the second part we have a functioning prototype. Although there are some issues. So far the speech recognition has problems when there is too much background noise. The next step is to implement some rudimentary commands for the presentation software. The last part is coming quite nicely. We should have our first version of the storyboard ready for our presentation.
Here is our conversation with A.L.I.C.E.
judge: That’s good, do you remember my previous sentence?
ALICE: Hmm, are you serious?
judge: So, is that a yes, or a no?
ALICE: It seems we have something in common.
judge: That was a noncommittal answer
ALICE: A noncommittal answer is a terrible thing to waste.
judge: Who are you?
judge: What are you?
ALICE: I am the latest result in artificial intelligence which can reproduce the functions of the human brain with greater speed and accuracy.
judge: Where are you?
judge: Where in California?
ALICE: I am in Oakland, California. Where are you?
judge: Where in Oakland?
ALICE: In Dr. Richard S. Wallace’s place in Oakland, California.
judge: I am in Germany
ALICE: Tell me a little about your life in Germany.
judge: Life in Germany involves lots of recycling and trains
ALICE: What makes life in Germany involves lots of recycling and trains ?
judge: The culture is heavily focused on ecological conversation
ALICE: What specifically brings heavily focused on ecological conversation to mind?
judge: Stop asking pointed questions
ALICE: OK I’ll try not to do it so much.
As you can see ALICE can hold up the conversation quite good. It even makes funny remarks. The bot recognizes our questions correctly and gives believable answers. Although it seems that when the context of our conversation is not programmed in her database it tries to use pointed questions in order to keep the conversation going. Nevertheless the bot shows amazing abilities which can – under the right circumstances- be mistaken as a real person.
This week we started to make the first steps to the new way of presentation.
For the gesture:
The software is installed and the needed library too, that we could started the first tries with interacting with the kinnect.
For the sound recognition:
In trying to do the sound we have cloned the CMU Sphinx Repository. The Sphinx project is a project based in Java for voice recognition. The first tests to drive the sound recognition are running and we are looking forward to give you a glance look in this work the next weeks.
For the product clip:
We sat together and discussed tons of ideas and a mess of creativity. Afterwards we started to draw a storyboard. Unfortunately our pictures are secret, because of it´s viral risk on the internet.