Locata dataset

CHANGELOG.rst

@@ -11,6 +11,13 @@ adheres to `Semantic Versioning <http://semver.org/spec/v2.0.0.html>`_.
 `Unreleased`_
 -------------
 
+
+Added
+~~~~~
+
+- Support for the `LOCATA <http://www.locata-challenge.org>`_ dataset in
+  ``pyroomacoustics.datasets.locata``
+
 Bugfix
 ~~~~~~
 
@@ -209,7 +216,6 @@ Added
 - Add iterative Wiener filtering approach for single channel denoising in
   ``pyroomacoustics.denoise.iterative_wiener``.
 
-
 Changed
 ~~~~~~~
 
@@ -302,6 +308,12 @@ Deprecation Notice
   ``pyroomacoustics.overlap_add``, etc, are now **deprecated**
   and will be removed in the near future
 
+Bugfix
+~~~~~~
+
+- Fixed the way multichannel signals are handled by
+  ``pyroomacoustics.datasets.AudioSample`` `plot` and `play` functions.
+
 `0.1.18`_ - 2018-04-24
 ----------------------
 

README.rst

@@ -107,6 +107,7 @@ moment we support the following.
 * `CMU ARCTIC <http://www.festvox.org/cmu_arctic/>`_
 * `TIMIT <https://catalog.ldc.upenn.edu/ldc93s1>`_
 * `Google Speech Commands Dataset <https://research.googleblog.com/2017/08/launching-speech-commands-dataset.html>`_
+* `LOCATA <http://www.locata-challenge.com>`_
 
 For more details, see the `doc <http://pyroomacoustics.readthedocs.io/en/pypi-release/pyroomacoustics.datasets.html>`_.
 

docs/pyroomacoustics.datasets.locata.rst

@@ -0,0 +1,7 @@
+pyroomacoustics.datasets.locata module
+======================================
+
+.. automodule:: pyroomacoustics.datasets.locata
+    :members:
+    :undoc-members:
+    :show-inheritance:

docs/pyroomacoustics.datasets.rst

@@ -16,6 +16,7 @@ Datasets Available
 
    pyroomacoustics.datasets.cmu_arctic
    pyroomacoustics.datasets.google_speech_commands
+   pyroomacoustics.datasets.locata
    pyroomacoustics.datasets.timit
 
 Tools and Helpers

examples/locata.py

@@ -0,0 +1,15 @@
+import numpy as np
+import pyroomacoustics as pra
+
+if __name__ == '__main__':
+
+    # parse arguments
+    parser = argparse.ArgumentParser(description='Test DOA algorithm on the locata data.')
+    parser.add_argument('-a', '--algo', choices=doa.algos.keys(),
+            help='doa algorithm')
+    parser.add_argument('-l', '--locata', type=str, default=None,
+            help='Location of LOCATA files')
+    parser.add_argument('--task', type=int, default=1,
+            help='LOCATA task number')
+    parser.add_argument('--rec', type=int, default=1,
+            help='LOCATA recording number')

examples/locata_challenge.py

@@ -0,0 +1,4 @@
+import argparse
+import pyroomacoustics as pra
+
+

pyroomacoustics/datasets/__init__.py

@@ -141,3 +141,4 @@
 from .timit import Word, Sentence, TimitCorpus
 from .cmu_arctic import CMUArcticCorpus, CMUArcticSentence, cmu_arctic_speakers
 from .google_speech_commands import GoogleSpeechCommands, GoogleSample
+from .locata import LOCATA, LocataRecording

pyroomacoustics/datasets/base.py

@@ -99,7 +99,7 @@ class AudioSample(Sample):
     The sampling frequency of the samples is an extra parameter.
 
     For multichannel audio, we assume the same format used by 
-    ```scipy.io.wavfile <https://docs.scipy.org/doc/scipy-0.14.0/reference/io.html#module-scipy.io.wavfile>`_``,
+    `scipy.io.wavfile <https://docs.scipy.org/doc/scipy-0.14.0/reference/io.html#module-scipy.io.wavfile>`_,
     that is ``data`` is then a 2D array with each column being a channel.
 
     Attributes
@@ -130,7 +130,12 @@ def play(self, **kwargs):
             print('Warning: sounddevice package is required to play audiofiles.')
             return
 
-        sd.play(self.data, samplerate=self.fs, **kwargs)
+        if self.data.ndim > 1:
+            data = self.data[:,0]
+        else:
+            data = self.data
+
+        sd.play(data, samplerate=self.fs, **kwargs)
 
     def plot(self, NFFT=512, noverlap=384, **kwargs):
         '''
@@ -150,6 +155,8 @@ def plot(self, NFFT=512, noverlap=384, **kwargs):
         # Handle single channel case
         if self.data.ndim == 1:
             data = self.data[:,None]
+        else:
+            data = self.data
 
         nchannels = data.shape[1]
 
@@ -158,11 +165,14 @@ def plot(self, NFFT=512, noverlap=384, **kwargs):
         prows = int(np.ceil(nchannels / pcols))
 
         for c in range(nchannels):
+            plt.subplot(prows, pcols, c+1)
             plt.specgram(data[:,c], NFFT=NFFT, Fs=self.fs, noverlap=noverlap, **kwargs)
             plt.xlabel('Time [s]')
             plt.ylabel('Frequency [Hz]')
             plt.title('Channel {}'.format(c+1))
 
+        plt.tight_layout(pad=0.5)
+
 
 class Dataset(object):
     '''

pyroomacoustics/datasets/cmu_arctic.py

@@ -33,12 +33,6 @@
 import numpy as np
 from scipy.io import wavfile
 
-try:
-    import sounddevice as sd
-    have_sounddevice = True
-except:
-    have_sounddevice = False
-
 from .utils import download_uncompress
 from .base import Meta, AudioSample, Dataset