Prior to presenting we were able to add the finishing touches to the piece which included slight adjustments to both the audio and the video. The audio was extended in length in order to make the piece less repetitive, this was done using material created by both Theo and myself. One issue that arose regarding the audio was that when cross fading between the distant and close tracks there would be a 3dB drop in volume when standing in the direct middle; this is due to logarithmic nature of sound pressure levels. In order to fix this we applied a shaped curve to the response from the Kinect to the audio such that the sound decayed logarithmically as the viewer left a certain quadrant. This was reasonably effective in fixing the problem however it would also be possible to increase the number of simultaneous tracks from four to nine, dividing the room into nine sections instead of quadrants. This would mean more crossover points in the audio but would result in a more intricate sound design; this is one of the ideas we plan to implement when installing the piece in Melbourne.
There were some difficulties encountered during the final day of presentation at MMU, which were handled with varying degrees of success. The first problem to arise was that the projector was accidentally knocked out of position about fifteen minutes before the beginning of the presentation. This meant that the mapping had to be done again; luckily it was done in time to begin. This suggests a problem with the project in that the projector was in reach of people accidentally moving it, had it happened during the presentation it would have been much more of a problem. A solution to this would be to mount to projector from the ceiling as to be out of the way. This would also solve the problem of shadows on the tower when people stand directly in front.
Another problem that we had was that our program was still unable to deal with multiple people in the frame causing the audio and video to jump between people or focus on people who were not engaged in the piece. During testing this was less of an issue because there were far less people around, ironically it worked best when there was no one to see it. This is a problem we are planning on fixing for the next installation by working code into our program which focuses on the nearest person or allows for multiple people in a different way. The final issue which we had not accounted for was that the ambient noise in the room was reasonable loud and made it difficult to interpret some of the audio content. Specifically at times when the audio was mainly bass, however the high end cut through well. A solution to this problem would simply to remove sections of the audio that do not have much high end frequency content, making the audio cut through ambient noise better.
Overall I believe the presentation was a good proof of concept. Some of the issues we were able to fix on the day and the others are all fixable for the Melbourne installation. The group worked together efficiently and effectively and there was an even output of work across the group which contributed to our finishing our work sooner than our schedule suggested we would. We expect this to continue with future work.