Closed tluocs closed 3 years ago
Thanx indeed there were some more short-term windows included in the mid-sized segments, in case short-term window length was > short-term window step (overlapping short-term frames case)
Fixed in https://github.com/tyiannak/pyAudioAnalysis/commit/f7f170daffb51a2f83eef0dd4a6c31e075b49dfe
It seems that the
mid_feature_extraction()
, as in theMidTermFeatures
module, contains an error in computing the ratio between the mid-term window and the short-term window. Details below.def mid_feature_extraction(signal, sampling_rate, mid_window, mid_step, short_window, short_step):
...
mid_window_ratio = int(round(mid_window / short_step))
I think the above line (Line 96) should be corrected to:
mid_window_ratio = round( (mid_window - (short_window - short_step)) / short_step )