source: python/lib/aubio/slicing.py @ f9400d0

feature/autosinkfeature/cnnfeature/cnn_orgfeature/constantqfeature/crepefeature/crepe_orgfeature/pitchshiftfeature/pydocstringsfeature/timestretchfix/ffmpeg5
Last change on this file since f9400d0 was f9400d0, checked in by Paul Brossier <piem@piem.org>, 6 years ago

[py] improve slice_source_at_stamps examples

  • Property mode set to 100644
File size: 6.2 KB
Line 
1"""utility routines to slice sound files at given timestamps"""
2
3import os
4from aubio import source, sink
5
6_max_timestamp = 1e120
7
8def slice_source_at_stamps(source_file, timestamps, timestamps_end=None,
9                           output_dir=None, samplerate=0, hopsize=256,
10                           create_first=False):
11    """Slice a sound file at given timestamps.
12
13    This function reads `source_file` and creates slices, new smaller
14    files each starting at `t` in `timestamps`, a list of integer
15    corresponding to time locations in `source_file`, in samples.
16
17    If `timestamps_end` is unspecified, the slices will end at
18    `timestamps_end[n] = timestamps[n+1]-1`, or the end of file.
19    Otherwise, `timestamps_end` should be a list with the same length
20    as `timestamps` containing the locations of the end of each slice.
21
22    If `output_dir` is unspecified, the new slices will be written in
23    the current directory. If `output_dir` is a string, new slices
24    will be written in `output_dir`, after creating the directory if
25    required.
26
27    The default `samplerate` is 0, meaning the original sampling rate
28    of `source_file` will be used. When using a sampling rate
29    different to the one of the original files, `timestamps` and
30    `timestamps_end` should be expressed in the re-sampled signal.
31
32    The `hopsize` parameter simply tells :class:`source` to use this
33    hopsize and does not change the output slices.
34
35    If `create_first` is True and `timestamps` does not start with `0`, the
36    first slice from `0` to `timestamps[0] - 1` will be automatically added.
37
38    Parameters
39    ----------
40    source_file : str
41        path of the resource to slice
42    timestamps : :obj:`list` of :obj:`int`
43        time stamps at which to slice, in samples
44    timestamps_end : :obj:`list` of :obj:`int` (optional)
45        time stamps at which to end the slices
46    output_dir : str (optional)
47        output directory to write the slices to
48    samplerate : int (optional)
49        samplerate to read the file at
50    hopsize : int (optional)
51        number of samples read from source per iteration
52    create_first : bool (optional)
53        always create the slice at the start of the file
54
55    Examples
56    --------
57    Create two slices: the first slice starts at the beginning of the
58    input file `loop.wav` and lasts exactly one second, starting at
59    sample `0` and ending at sample `44099`; the second slice starts
60    at sample `44100` and lasts until the end of the input file:
61
62    >>> aubio.slice_source_at_stamps('loop.wav', [0, 44100])
63
64    Create one slice, from 1 second to 2 seconds:
65
66    >>> aubio.slice_source_at_stamps('loop.wav', [44100], [44100 * 2 - 1])
67
68    Notes
69    -----
70    Slices may be overlapping. If `timestamps_end` is `1` element
71    shorter than `timestamps`, the last slice will end at the end of
72    the file.
73    """
74
75    if timestamps is None or len(timestamps) == 0:
76        raise ValueError("no timestamps given")
77
78    if timestamps[0] != 0 and create_first:
79        timestamps = [0] + timestamps
80        if timestamps_end is not None:
81            timestamps_end = [timestamps[1] - 1] + timestamps_end
82
83    if timestamps_end is not None:
84        if len(timestamps_end) == len(timestamps) - 1:
85            timestamps_end = timestamps_end + [_max_timestamp]
86        elif len(timestamps_end) != len(timestamps):
87            raise ValueError("len(timestamps_end) != len(timestamps)")
88    else:
89        timestamps_end = [t - 1 for t in timestamps[1:]] + [_max_timestamp]
90
91    regions = list(zip(timestamps, timestamps_end))
92    #print regions
93
94    source_base_name, _ = os.path.splitext(os.path.basename(source_file))
95    if output_dir is not None:
96        if not os.path.isdir(output_dir):
97            os.makedirs(output_dir)
98        source_base_name = os.path.join(output_dir, source_base_name)
99
100    def new_sink_name(source_base_name, timestamp, samplerate):
101        """ create a sink based on a timestamp in samples, converted in seconds """
102        timestamp_seconds = timestamp / float(samplerate)
103        return source_base_name + "_%011.6f" % timestamp_seconds + '.wav'
104
105    # open source file
106    _source = source(source_file, samplerate, hopsize)
107    samplerate = _source.samplerate
108
109    total_frames = 0
110    slices = []
111
112    while True:
113        # get hopsize new samples from source
114        vec, read = _source.do_multi()
115        # if the total number of frames read will exceed the next region start
116        while len(regions) and total_frames + read >= regions[0][0]:
117            #print "getting", regions[0], "at", total_frames
118            # get next region
119            start_stamp, end_stamp = regions.pop(0)
120            # create a name for the sink
121            new_sink_path = new_sink_name(source_base_name, start_stamp, samplerate)
122            # create its sink
123            _sink = sink(new_sink_path, samplerate, _source.channels)
124            # create a dictionary containing all this
125            new_slice = {'start_stamp': start_stamp, 'end_stamp': end_stamp, 'sink': _sink}
126            # append the dictionary to the current list of slices
127            slices.append(new_slice)
128
129        for current_slice in slices:
130            start_stamp = current_slice['start_stamp']
131            end_stamp = current_slice['end_stamp']
132            _sink = current_slice['sink']
133            # sample index to start writing from new source vector
134            start = max(start_stamp - total_frames, 0)
135            # number of samples yet to written be until end of region
136            remaining = end_stamp - total_frames + 1
137            #print current_slice, remaining, start
138            # not enough frames remaining, time to split
139            if remaining < read:
140                if remaining > start:
141                    # write remaining samples from current region
142                    _sink.do_multi(vec[:, start:remaining], remaining - start)
143                    #print("closing region", "remaining", remaining)
144                    # close this file
145                    _sink.close()
146            elif read > start:
147                # write all the samples
148                _sink.do_multi(vec[:, start:read], read - start)
149        total_frames += read
150        # remove old slices
151        slices = list(filter(lambda s: s['end_stamp'] > total_frames,
152            slices))
153        if read < hopsize:
154            break
Note: See TracBrowser for help on using the repository browser.