The Hood Game is my most recent work at CADLab. The idea began as a study a number of years ago regarding the control people with dysarthria have over their voice. Dysarthria is a speaking disorder brought on by neurological damage. People can be born with the disorder such as the case with Cerebral Palsy or can develop the disorder over time from things such as Parkinson's Disease or induced by a stroke. The level to which speech is affected varies from individual to individual. The result is the same, the person retains the ability to create vocal noises but loose varying degrees of control over their speech.
The Hood Study looked at children with dysarthria to see if they retained the basic building blocks of speech, prosody. Prosody is made up the pitch, intensity and duration of a vocalization. These are the foundation of generating controlled speech. The study concluded that even young children with dysarthria can still control the prosody of their voice which was rather surprising.
The question that arose from this study is what happens if people with dysarthria practice controlling these basic foundations of their speech, will it have a linear improvement on their speech production.
The game was written entirely by myself in Java in about 2.5 months. To date there are around 20,000 lines of code. One of the big challenges of this game was the original study used a Wizard of Oz test. Children were asked to generate for example a short, mid, or long duration sound prompted by the computer and their parent in another room had to guess which duration measure it was. As far as the child was concerned, the computer was scoring the game when in fact it was dependent upon whether the parent scored correctly or incorrectly. This game obviously did not have that luxury and the first thing to be written was a real time pitch extractor in Java which can be read about further here.
The game is also designed to be downloaded to people's home computers which again adds a level of complexity not seen in the original study. Results of progress are now encrypted and sent back to a custom built server written in Java using Diffie-Helman and perfect forward secrecy. The microphone volume is automatically calibrated, the games are animated and there are a host of other features.
Figure 2. This screen shot shows part of the automated microphone calibration. Calibration is handled in two parts, asking for silence and then for a sustained "Ahhh" sound to be made. Though hard to see in this screen shot, a glass JPanel is used during this process to capture all keyboard and mouse events. This is used to prevent the user from changing screens while the automated process is running.
Figure 3. This screen shot shows the last step of microphone calibration. The user can test the final volume adjustment visually or if the auto-calibration failed (which is bound to happen eventually), they can manually adjust the volume until the visual response shows appropriate values.
Figure 4. This screen shows the duration game. There are three space ships and as the game progresses, a UFO will appear on the right hand side to signify which duration sound to make. A short, medium or long duration sound. A correct guess will result in a laser being fired from a spaceship to blow up the UFO otherwise the UFO will escape.
Firgure 4. This screen shows the same duration game but in a training feature. A visual bar across the screen shows what the desired length of sound to produce. As the sound is made, a red bar progresses across to give visual feedback on when to stop. As the level increases, the variance from the correct length of sound is shortened and thus the visual bar will also be shortened.
Figure 5. This game shows the pitch game. The pitch game works exactly as the duration game but the average pitch over a sustained time is used instead of the duration of the sound. Produce an appropriate pitch and the fish eats the worm, miss and the worm escapes.