Into the last week of the capstone project, we managed to further reduce the false positive rate of intention recognition. One more user test is done with our new user Eryn, whose game session successfully passed all the usability tests. The session is recorded in the final demo video. The team is currently working on the final poster and expects to finish both the poster and report by next week.
This week I helped with the integration testing and recorded the final demo video. Fixed some bugs in the game statistics display. Contributed to the final poster. Next week’s effort will focus on the final report.
This week I gave the final presentation and received Prof. Sullivan’s feedback regarding the expressiveness of correlation coefficients for the fidelity of the filtered signal. I will address this point in our final report. I also helped with the test with Eryn. It is exciting that she passed the user test. Finally, I helped with making the final poster. Next week I will focus on writing up the final report.
This week I worked on improving the reliability and stability of our neural network to reduce false positive rate in our game. From comprehensive testing with our newly acquired test subject Eryn, we found 40% as a threshold for firing provides the best result for fire control, as confirmed by visual examination. I also worked with Lavender and Eryn to obtain our last test video.
This week our group is working on preparing for the final presentation.
On the software side, we made necessary changes to the code for effective latency tests under running conditions.
On the game logic side, we are improving the user interface and creating a more user friendly game experience by adding necessary components such as score display and start-over buttons.
This week the game implementation was finalized. I further improved the UI by adding more design details like scoring display, the start-over button, device connectivity status display, guiding instructions for calibration phase, option to save user brain profile, testing playground and etc. Next week I’ll just help the team with the final presentation.
This week I helped Lavender prepare for her final presentation. I also wrote code in our program to test the latency of each stage of our algorithm. Currently, we can profile for each stage (signal processing, NN inference, rendering) except the data gathering. We tried to find a reliable way to synchronize the data collection with a timer, but we can’t find a way to reliably time it. Tomorrow we will be testing each stage in conjunction with the profiler and see if latency at each stage change significantly.