Winter Week 8 - More Testing

This week we spent our lab time fine-tuning our lowpass filter for noise and finalizing our output file format. We realized that placing the cutoff frequency of the filter too low took out too much information for our recognition algorithm, as some chords were incorrectly characterized. We moved the filter up by 1000 Hz and tested again, continuing this process until we were satisfied with a 6000 Hz cutoff.

Next, we figured out a way to print chord information to a file in a precise format. We tried to use the C clock class to get accurate timing information, but couldn't get the code to execute properly for some reason. Instead, we estimated a clock by using our sample rate and frame size. Our output file consists of the chord being played with a timestamp of the interval it was played for, followed by each subsequent chord. If no chord is played during an interval, it is labeled as "NC" for no chord.

Lastly, we began testing with chord sequences, where we played a sequence of ten chords at regular intervals. We found that matching the tempo matters a lot, so having the metronome play will be the key to the success of the chord recognition. Next week, we will continue to test chord sequences and then begin our final presentation and report.