source: python/demos/demo_tempo_plot.py @ 00fcc5d

feature/autosinkfeature/cnnfeature/cnn_orgfeature/constantqfeature/crepefeature/crepe_orgfeature/pitchshiftfeature/timestretchfix/ffmpeg5
Last change on this file since 00fcc5d was 4120fbc, checked in by Paul Brossier <piem@piem.org>, 8 years ago

python/demos: python3 and double precision compatibility

  • Property mode set to 100755
File size: 2.4 KB
Line 
1#! /usr/bin/env python
2
3import sys
4from aubio import tempo, source
5
6win_s = 512                 # fft size
7hop_s = win_s // 2          # hop size
8
9if len(sys.argv) < 2:
10    print("Usage: %s <filename> [samplerate]" % sys.argv[0])
11    sys.exit(1)
12
13filename = sys.argv[1]
14
15samplerate = 0
16if len( sys.argv ) > 2: samplerate = int(sys.argv[2])
17
18s = source(filename, samplerate, hop_s)
19samplerate = s.samplerate
20o = tempo("default", win_s, hop_s, samplerate)
21
22# tempo detection delay, in samples
23# default to 4 blocks delay to catch up with
24delay = 4. * hop_s
25
26# list of beats, in samples
27beats = []
28
29# total number of frames read
30total_frames = 0
31while True:
32    samples, read = s()
33    is_beat = o(samples)
34    if is_beat:
35        this_beat = o.get_last_s()
36        beats.append(this_beat)
37    total_frames += read
38    if read < hop_s: break
39
40if len(beats) > 1:
41    # do plotting
42    from numpy import mean, median, diff
43    import matplotlib.pyplot as plt
44    bpms = 60./ diff(beats)
45    print('mean period: %.2fbpm, median: %.2fbpm' % (mean(bpms), median(bpms)))
46    print('plotting %s' % filename)
47    plt1 = plt.axes([0.1, 0.75, 0.8, 0.19])
48    plt2 = plt.axes([0.1, 0.1, 0.8, 0.65], sharex = plt1)
49    plt.rc('lines',linewidth='.8')
50    for stamp in beats: plt1.plot([stamp, stamp], [-1., 1.], '-r')
51    plt1.axis(xmin = 0., xmax = total_frames / float(samplerate) )
52    plt1.xaxis.set_visible(False)
53    plt1.yaxis.set_visible(False)
54
55    # plot actual periods
56    plt2.plot(beats[1:], bpms, '-', label = 'raw')
57
58    # plot moving median of 5 last periods
59    median_win_s = 5
60    bpms_median = [ median(bpms[i:i + median_win_s:1]) for i in range(len(bpms) - median_win_s ) ]
61    plt2.plot(beats[median_win_s+1:], bpms_median, '-', label = 'median of %d' % median_win_s)
62    # plot moving median of 10 last periods
63    median_win_s = 20
64    bpms_median = [ median(bpms[i:i + median_win_s:1]) for i in range(len(bpms) - median_win_s ) ]
65    plt2.plot(beats[median_win_s+1:], bpms_median, '-', label = 'median of %d' % median_win_s)
66
67    plt2.axis(ymin = min(bpms), ymax = max(bpms))
68    #plt2.axis(ymin = 40, ymax = 240)
69    plt.xlabel('time (mm:ss)')
70    plt.ylabel('beats per minute (bpm)')
71    plt2.set_xticklabels([ "%02d:%02d" % (t/60, t%60) for t in plt2.get_xticks()[:-1]], rotation = 50)
72
73    #plt.savefig('/tmp/t.png', dpi=200)
74    plt2.legend()
75    plt.show()
76
77else:
78    print('mean period: %.2fbpm, median: %.2fbpm' % (0, 0))
79    print('plotting %s' % filename)
Note: See TracBrowser for help on using the repository browser.