Source code for

#!/usr/bin/env python
# ----------------------------------------------------------------------
# Filename:
#  Purpose: Python Class for transforming seismograms to audio WAV files
#   Author: Moritz Beyreuther
#    Email:
# Copyright (C) 2009-2012 Moritz Beyreuther
# ------------------------------------------------------------------------
WAV bindings to ObsPy core module.

    The ObsPy Development Team (
    GNU Lesser General Public License, Version 3
from pathlib import Path
import wave

import numpy as np

from obspy import Stream, Trace
from obspy.core.compatibility import from_buffer

# WAVE data format is unsigned char up to 8bit, and signed int
# for the remaining.
    1: '<u1',  # unsigned char
    2: '<i2',  # signed short int
    4: '<i4',  # signed int (int32)

[docs]def _is_wav(filename): """ Checks whether a file is a audio WAV file or not. :type filename: str :param filename: Name of the audio WAV file to be checked. :rtype: bool :return: ``True`` if a WAV file. .. rubric:: Example >>> _is_wav("/path/to/3cssan.near.8.1.RNON.wav") #doctest: +SKIP True """ try: fh =, 'rb') try: (_nchannel, width, _rate, _len, _comptype, _compname) = \ fh.getparams() finally: fh.close() except Exception: return False if width in [1, 2, 4]: return True return False
[docs]def _read_wav(filename, headonly=False, **kwargs): # @UnusedVariable """ Reads a audio WAV file and returns an ObsPy Stream object. Currently supports uncompressed unsigned char and short integer and integer data values. This should cover most WAV files. .. warning:: This function should NOT be called directly, it registers via the ObsPy :func:`` function, call this instead. :type filename: str :param filename: Audio WAV file to be read. :rtype: :class:`` :return: A ObsPy Stream object. .. rubric:: Example >>> from obspy import read >>> st = read("/path/to/3cssan.near.8.1.RNON.wav") >>> print(st) #doctest: +NORMALIZE_WHITESPACE 1 Trace(s) in Stream: ... | 1970-01-01T00:00:00.000000Z - 1970-01-01T00:00:00.371143Z | 7000.0 Hz, 2599 samples """ # read WAV file fh =, 'rb') try: # header information (_nchannel, width, rate, length, _comptype, _compname) = fh.getparams() header = {'sampling_rate': rate, 'npts': length} if headonly: return Stream([Trace(header=header)]) if width not in WIDTH2DTYPE.keys(): msg = "Unsupported Format Type, word width %dbytes" % width raise TypeError(msg) data = from_buffer(fh.readframes(length), dtype=WIDTH2DTYPE[width]) finally: fh.close() return Stream([Trace(header=header, data=data)])
[docs]def _write_wav(stream, filename, framerate=7000, rescale=False, width=None, **kwargs): # @UnusedVariable """ Writes a audio WAV file from given ObsPy Stream object. The seismogram is squeezed to audible frequencies. The generated WAV sound file is as a result really short. The data are written uncompressed as signed 4-byte integers. .. warning:: This function should NOT be called directly, it registers via the the :meth:`` method of an ObsPy :class:`` object, call this instead. :type stream: :class:`` :param stream: The ObsPy Stream object to write. :type filename: str :param filename: Name of the audio WAV file to write. :type framerate: int, optional :param framerate: Sample rate of WAV file to use. This this will squeeze the seismogram (default is 7000). :type rescale: bool, optional :param rescale: Maximum to maximal representable number :type width: int, optional :param width: dtype to write, 1 for '<u1', 2 for '<i2' or 4 for '<i4'. tries to autodetect width from data, uses 4 otherwise """ i = 0 file_path = Path(filename) base = file_path.parent / file_path.stem if width not in WIDTH2DTYPE.keys() and width is not None: raise TypeError("Unsupported Format Type, word width %dbytes" % width) for trace in stream: # try to autodetect width from data, see #791 if width is None: if[-2:] in ['u1', 'i2', 'i4']: tr_width = int([-1]) else: tr_width = 4 else: tr_width = width # write WAV file if len(stream) >= 2: filename = "%s%03d%s" % (base, i, file_path.suffix) w =, 'wb') try: trace.stats.npts = len( # (nchannels, sampwidth, framerate, nframes, comptype, compname) w.setparams((1, tr_width, framerate, trace.stats.npts, 'NONE', 'not compressed')) data = dtype = WIDTH2DTYPE[tr_width] if rescale: # optimal scale, account for +/- and the zero maxint = 2 ** (tr_width * 8 - 1) - 1 # upcast for following rescaling data = data.astype(np.float64) data = data / abs(data).max() * maxint data = np.require(data, dtype=dtype) w.writeframes(data.tobytes()) finally: w.close() i += 1
if __name__ == '__main__': import doctest doctest.testmod(exclude_empty=True)