Merge pull request #63 from SuperKogito/fix-cqcc-shape

fix cqcc size bug

Author: SuperKogito

build.sh

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/fbanks/bark_fbanks.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional, Tuple
 
 import numpy as np

spafe/fbanks/gammatone_fbanks.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/fbanks/linear_fbanks.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/fbanks/mel_fbanks.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/bfcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np
@@ -166,7 +167,7 @@ def cqcc(
     normalize: Optional[NormalizationType] = None,
     number_of_octaves: int = 7,
     number_of_bins_per_octave: int = 24,
-    resampling_ratio: float = 0.95,
+    resampling_ratio: float = 1.0,
     spectral_threshold: float = 0.005,
     f0: float = 120,
     q_rate: float = 1.0,
@@ -208,7 +209,7 @@ def cqcc(
         number_of_bins_per_octave   (int) : numbers of bins oer occtave.
                                             (Default is 24).
         resampling_ratio          (float) : ratio to use for the uniform resampling.
-                                            (Default is 0.95).
+                                            (Default is 1.00).
         spectral_threshold        (float) : spectral threshold.
                                             (Default is 0.005).
         f0                        (float) : fundamental frequency.
@@ -217,7 +218,7 @@ def cqcc(
                                             (Default is 1.0).
 
     Returns:
-        (numpy.ndarray) : 2d array of BFCC features (num_frames x num_ceps).
+        (numpy.ndarray) : 2d array of BFCC features (num_frames*resampling_ratio x num_ceps).
 
     Tip:
         - :code:`dct` : can take the following options [1, 2, 3, 4].
@@ -283,7 +284,7 @@ def cqcc(
     # -> log(.)
     # handle zeros: if feat is zero, we get problems with log
     features_no_zero = zero_handling(x=power_spectrum)
-    log_features = np.log(features_no_zero)
+    log_features = np.log(features_no_zero.T)
 
     # uniform resampling
     resampled_features = resample(
@@ -300,4 +301,5 @@ def cqcc(
     # normalization
     if normalize:
         cqccs = normalize_ceps(cqccs, normalize)
+
     return cqccs

spafe/features/cqcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional, Tuple
 
 import numpy as np

spafe/features/gfcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/lfcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/lpc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/mfcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/msrcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/ngcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/pncc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np

spafe/features/psrcc.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import Optional
 
 import numpy as np
@@ -148,9 +149,7 @@ def __rastaplp(
 
     # equal loudness pre_emphasis
     E = lambda w: ((w**2 + 56.8 * 10**6) * w**4) / (
-        (w**2 + 6.3 * 10**6)
-        * (w**2 + 0.38 * 10**9)
-        * (w**6 + 9.58 * 10**26)
+        (w**2 + 6.3 * 10**6) * (w**2 + 0.38 * 10**9) * (w**6 + 9.58 * 10**26)
     )
     Y = [E(w) for w in auditory_spectrum]
 

spafe/features/rplp.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 import numpy as np
 from scipy.signal import lfilter
 from typing_extensions import Literal

spafe/utils/cepstral.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 import numpy as np
 from typing_extensions import Literal
 

spafe/utils/converters.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 import numpy as np
 from scipy import signal
 from typing_extensions import Literal

spafe/utils/filters.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 from typing import List
 
 import numpy as np

spafe/utils/spectral.py

@@ -6,6 +6,7 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 import numpy as np
 
 from spafe.utils.converters import hz2mel, hz2bark, hz2erb

spafe/utils/vis.py

@@ -6,4 +6,5 @@
   For a copy, see <https://github.com/SuperKogito/spafe/blob/master/LICENSE>.
 
 """
+
 __version__ = "0.3.2"

spafe/version.py

@@ -3,6 +3,7 @@
 import scipy.io.wavfile
 from spafe.utils import vis
 from spafe.features.cqcc import cqcc
+from spafe.utils.preprocessing import framing
 from spafe.utils.exceptions import ParameterError
 from spafe.utils.cepstral import normalize_ceps, lifter_ceps
 
@@ -18,6 +19,7 @@
 @pytest.mark.parametrize("dct_type", [1, 2, 4])
 @pytest.mark.parametrize("lifter", [None, 0.7, -7])
 @pytest.mark.parametrize("normalize", [None, "mvn", "ms"])
+@pytest.mark.parametrize("resampling_ratio", [1.0, 0.9, 0.3])
 def test_cqcc(
     sig,
     fs,
@@ -29,6 +31,7 @@ def test_cqcc(
     dct_type,
     lifter,
     normalize,
+    resampling_ratio,
 ):
     """
     test cqcc features module for the following:
@@ -40,6 +43,9 @@ def test_cqcc(
         - check normalization.
         - check liftering.
     """
+    # get number of frames
+    frames, frame_length = framing(sig=sig, fs=fs, win_len=0.025, win_hop=0.01)
+    num_frames = len(frames)
 
     # check error for number of filters is smaller than number of cepstrums
     with pytest.raises(ParameterError):
@@ -50,6 +56,7 @@ def test_cqcc(
             nfft=nfft,
             low_freq=low_freq,
             high_freq=fs,
+            resampling_ratio=resampling_ratio,
         )
 
     # compute features
@@ -64,12 +71,17 @@ def test_cqcc(
         dct_type=dct_type,
         lifter=lifter,
         normalize=normalize,
+        resampling_ratio=resampling_ratio,
     )
 
     # assert number of returned cepstrum coefficients
     if not cqccs.shape[1] == num_ceps:
         raise AssertionError
 
+    # assert number of returned cepstrum coefficients
+    if not cqccs.shape[0] == int(num_frames * resampling_ratio):
+        raise AssertionError
+
     # check normalize
     if normalize:
         np.testing.assert_array_almost_equal(
@@ -86,6 +98,7 @@ def test_cqcc(
                     dct_type=dct_type,
                     lifter=lifter,
                     normalize=None,
+                    resampling_ratio=resampling_ratio,
                 ),
                 normalize,
             ),
@@ -108,6 +121,7 @@ def test_cqcc(
                         dct_type=dct_type,
                         lifter=None,
                         normalize=normalize,
+                        resampling_ratio=resampling_ratio,
                     ),
                     lifter,
                 ),