Closed daccordeon closed 4 years ago
initial time series generated, tracker jumps in one ring, then back out
more videos taken and tracked overall, tracking is very poor at high frequencies, and requires very slow and low amplitude oscillations note: frequency resolution is fps/total_frames expt_1.pdf expt_2.pdf expt_3.pdf
still to do:
also:
we could also try recording the intensity at the centre of the interference pattern to see if this can get high frequencies better than tracking
centre measurement provides similar results at low frequencies at high frequencies is unclear if significantly better than tracker note: likely that human driven high frequencies are very inconsistent, which might explain dips in signal and lack of signal in spectrum
expt_0_centre.pdf expt_3_centre.pdf
going to the labs:
eventually:
edit: formatting and edit tag
adding in budget webcam and speaker driven by function generator produces very sharp results note: expt_4_xxyy, driving frequency to speaker is xx.yy Hz
expt_4_0209_centre.pdf expt_4_0580_centre.pdf expt_4_2014_centre.pdf
first milestone by end of tomorrow:
edit: formatting
viterbi successful on long 5/10 minute captures scanning_range, the amount that the viterbi path is allowed to change frequency in one long timestep is better left low at around 3-5 bins
expt_ngrid_5_slow.pdf expt_viterbi_path_5_slow.pdf
all that's left for this issue:
edit: formatting
also, although this is likely hard:
see experiment_viterbi.py and .ipynb
closing this issue, open a new issue for further extensions