def cluster(src, key="AvgTonalCentroid(6)", nbins=15, min_volume=-30):
jsonfile = "%s-%s-%d.json" % (src, key, nbins)
if os.path.exists(jsonfile):
return deserialize(jsonfile)
analysis = meap.analysis(src)
segs = []
features = []
rms = []
for idx, X in enumerate(analysis):
segs.append((X["onset_time"], X["chunk_length"], idx))
features.append(X[key])
rms.append(X["RMSAmplitude(1)"])
rms = np.array(rms)
features = np.array(features)
segs = np.array(segs)
too_quiet = rms < min_volume
quiet_segs = segs[too_quiet]
features = features[np.invert(too_quiet)]
segs = segs[np.invert(too_quiet)]
codebook, distortion = kmeans(features, nbins)
clusters, distance = vq(features, codebook)
out = {}
for i in range(nbins + 1):
out[i] = []
for idx, cluster in enumerate(clusters):
out[cluster].append(segs[idx].tolist())
out[nbins] = quiet_segs.tolist()
serialize(out, jsonfile)
return out
def cluster(src, key="AvgTonalCentroid(6)", nbins=15, min_volume=-30):
jsonfile = "%s-%s-%d.json" % (src, key, nbins)
if os.path.exists(jsonfile):
return deserialize(jsonfile)
analysis = meap.analysis(src)
segs = []
features = []
rms = []
for idx, X in enumerate(analysis):
segs.append((X["onset_time"], X["chunk_length"], idx))
features.append(X[key])
rms.append(X["RMSAmplitude(1)"])
rms = np.array(rms)
features = np.array(features)
segs = np.array(segs)
too_quiet = rms < min_volume
quiet_segs = segs[too_quiet]
features = features[np.invert(too_quiet)]
segs = segs[np.invert(too_quiet)]
codebook, distortion = kmeans(features, nbins)
clusters, distance = vq(features, codebook)
out = {}
for i in range(nbins + 1):
out[i] = []
for idx, cluster in enumerate(clusters):
out[cluster].append(segs[idx].tolist())
out[nbins] = quiet_segs.tolist()
serialize(out, jsonfile)
return out
def _get_features(self):
a = meap.analysis(self.path)
out = [X[self.key] for X in a]
return np.array(out)
def _get_segments(self):
a = meap.analysis(self.path)
return [Seg(X["onset_time"], X["chunk_length"], idx)
for idx,X in enumerate(a)]
def _get_features(self):
a = meap.analysis(self.path)
out = [X[self.key] for X in a]
return np.array(out)
def _get_segments(self):
a = meap.analysis(self.path)
return [
Seg(X["onset_time"], X["chunk_length"], idx)
for idx, X in enumerate(a)
]
