Merge bc1ba08 into fb87932

librosa/effects.py

@@ -33,12 +33,13 @@
 
 from . import core
 from . import decompose
+from . import feature
 from . import util
 from .util.exceptions import ParameterError
 
 __all__ = ['hpss', 'harmonic', 'percussive',
            'time_stretch', 'pitch_shift',
-           'remix']
+           'remix', 'trim']
 
 
 def hpss(y, **kwargs):
@@ -77,7 +78,7 @@ def hpss(y, **kwargs):
     >>> y, sr = librosa.load(librosa.util.example_audio_file())
     >>> y_harmonic, y_percussive = librosa.effects.hpss(y)
 
-    >>> # Get a more isolated percussive component by widening its margin 
+    >>> # Get a more isolated percussive component by widening its margin
     >>> y_harmonic, y_percussive = librosa.effects.hpss(y, margin=(1.0,5.0))
 
     '''
@@ -94,6 +95,7 @@ def hpss(y, **kwargs):
 
     return y_harm, y_perc
 
+
 def harmonic(y, **kwargs):
     '''Extract harmonic elements from an audio time-series.
 
@@ -117,14 +119,13 @@ def harmonic(y, **kwargs):
 
     Examples
     --------
-    >>> # Extract harmonic component 
+    >>> # Extract harmonic component
     >>> y, sr = librosa.load(librosa.util.example_audio_file())
     >>> y_harmonic = librosa.effects.harmonic(y)
 
     >>> # Use a margin > 1.0 for greater harmonic separation
     >>> y_harmonic = librosa.effects.harmonic(y, margin=3.0)
 
-
     '''
 
     # Compute the STFT matrix
@@ -138,6 +139,7 @@ def harmonic(y, **kwargs):
 
     return y_harm
 
+
 def percussive(y, **kwargs):
     '''Extract percussive elements from an audio time-series.
 
@@ -160,15 +162,14 @@ def percussive(y, **kwargs):
     librosa.decompose.hpss : HPSS for spectrograms
 
     Examples
-    --------    
-    >>> # Extract percussive component 
+    --------
+    >>> # Extract percussive component
     >>> y, sr = librosa.load(librosa.util.example_audio_file())
     >>> y_percussive = librosa.effects.percussive(y)
 
     >>> # Use a margin > 1.0 for greater percussive separation
     >>> y_percussive = librosa.effects.percussive(y, margin=3.0)
 
-
     '''
 
     # Compute the STFT matrix
@@ -374,3 +375,74 @@ def remix(y, intervals, align_zeros=True):
         y_out.append(y[clip])
 
     return np.concatenate(y_out, axis=-1)
+
+
+def trim(y, top_db=60, n_fft=2048, hop_length=512, index=False):
+    '''Trim leading and trailing silence from an audio signal.
+
+    Parameters
+    ----------
+    y : np.ndarray, shape=(n,) or (2,n)
+        Audio signal, can be mono or stereo
+
+    top_db : number > 0
+        The threshold (in decibels) below peak to consider as
+        silence
+
+    n_fft : int > 0
+        The number of samples per analysis frame
+
+    hop_length : int > 0
+        The number of samples between analysis frames
+
+    index : bool
+        If `True`, return the start and end of the non-silent
+        region of `y` along with the trimmed signal.
+
+        If `False`, only return the trimmed signal.
+
+    Returns
+    -------
+    y_trimmed : np.ndarray, shape=(m,) or (2, m)
+        The trimmed signal
+
+    index : slice, optional
+        If `index=True` is provided, then this contains
+        the slice of `y` corresponding to the non-silent region:
+        `y_trimmed = y[index]`.
+
+    Examples
+    --------
+    >>> # Load some audio
+    >>> y, sr = librosa.load(librosa.util.example_audio_file())
+    >>> # Trim the beginning and ending silence
+    >>> yt = librosa.effects.trim(y)
+    >>> # Print the durations
+    >>> print(librosa.get_duration(y), librosa.get_duration(yt))
+    61.45886621315193 60.58086167800454
+    '''
+
+    # Convert to mono
+    y_mono = core.to_mono(y)
+
+    # Compute the MSE for the signal
+    mse = feature.rmse(y=y_mono, n_fft=n_fft, hop_length=hop_length)**2
+
+    # Compute the log power indicator and non-zero positions
+    logp = core.logamplitude(mse, ref_power=np.max, top_db=None) > - top_db
+    nonzero = np.flatnonzero(logp)
+
+    # Compute the start and end positions
+    # End position goes one frame past the last non-zero
+    start = int(core.frames_to_samples(nonzero[0], hop_length))
+    end = min(len(y_mono),
+              int(core.frames_to_samples(nonzero[-1] + 1, hop_length)))
+
+    # Build the mono/stereo index
+    full_index = [slice(None)] * y.ndim
+    full_index[-1] = slice(start, end)
+
+    if index:
+        return y[full_index], full_index[-1]
+    else:
+        return y[full_index]